open/ms-ai-architect
1
0
Fork 0
Code Issues Releases 1 Activity

fix(ms-ai-architect): de 17 ratifiserte er anvendt — og «promotering» matte defineres foer den kunne utfoeres

Browse source 
#21-#24 ble ratifisert i 39c947a, men ratifisering er ikke anvendelse. Denne
oekten gjoer de 17 editene og lukker entriene. 86 linjer slettet, 7 satt inn,
over 17 filer i 4 skills. Suite 1052/1052. Koe: 53 entries (15 aapne, 38 resolved).

PROMOTERING FANTES IKKE FRA FOER. #22 sier «PROMOTE any surviving keep-class
line to the enclosing level», men ingen av de 21 resolved entriene hadde en
survivor — alle 40 tidligere editer var rene slettinger. Maalt foer valget, ikke
antatt. Promotering er derfor definert her: fjern listemarkoeren, slik at linjen
blir et frittstaaende avsnitt der etiketten stod. AA la survivors staa som
bullets ville i idx-26x gjort dem til element fire og fem i lista rett over —
referent-drift, som er noeyaktig det #22 finnes for aa hindre. AA fjerne «- »
forfatter null ord, og er derfor tillatt under NEVER RENAME.

STRENG-INVARIANTEN SAA IKKE DEN ANDRE DEFEKTEN. To promoterte linjer uten blank
linje mellom seg er ETT markdown-avsnitt: «25+ unike Microsoft Learn URLs
referert» og «15+ PowerShell code samples inkludert» smeltet til én paastand.
Hel-linje-sjekken passerte baade foer og etter rettelsen — den ser tilstedevaerelse,
aldri koherens. Fanget ved aa lese den gjengitte formen.

VERIFISERT, IKKE PAASTAATT: 65 slettelinjer maalt fravaerende og 20 keep-linjer
maalt tilstede noeyaktig én gang, ved hel-linje-sammenligning. Foerste
gjennomkjoering brukte fragmenter og ga to falske positiver («Azure OpenAI
reasoning models», «manage-costs») som begge stod legitimt i kildelister —
substring-match var feil invariant, ikke korpus.

RESIDU-SVEIPET BETALTE FOR TRETTENDE GANG. Ved siden av editene ligger en
uratifisert klasse: per-paastands provenance-markoerer som navngir verktoey, altsaa
#22s slette-referent, men utenfor de 58 og utenfor alle ankere. Maalt korpus-bredt:
18 «**Confidence:** Verified (microsoft-learn docs_*)» + 11 «*Verified via
microsoft_docs_*» + 13 inline + 1 gjenlevende tidsstempel. 28 av ~42 staar i ÉN
fil — den idx-26y nettopp redigerte. Bokfoert som idx-26as og bevisst IKKE
arvet, paa samme grunn som idx-26ar: de henger paa en SPESIFIKK paastand, ikke paa
dokumentet, saa sletting fjerner ogsaa et konfidenssignal footer-klassen aldri
hadde. Den asymmetrien er hvorfor dette er et spoersmaal og ikke en arv.

Et grep-treff til ble talt foer det ble klassifisert: «**Verifiseringsdato:**
YYYY-MM-DD» er en mal-plassholder, ikke en forekomst. Det holdt klassen paa 1.
This commit is contained in:
Kjell Tore Guttormsen 2026-08-12 00:11:03 +02:00
commit 0a3ce4f414
18 changed files with 91 additions and 120 deletions
Download patch file Download diff file Expand all files Collapse all files

117
scripts/kb-eval/data/g7-review-queue.json
View file

@ -409,19 +409,20 @@
"id": "idx-26w",
"file": "skills/ms-ai-engineering/references/agent-orchestration/agent-to-agent-communication.md",
"class": "replacement",
"status": "open",
"status": "resolved",
"raised": "2026-08-09",
"summary": "Booked while closing idx-26o, from the same what-remains check that produced idx-26t. '**Total sources cited**: 7 unique URLs fra MCP-research' passes as ARITHMETIC and fails as PROVENANCE - measured, not inferred. The count is right: 7 numbered entries, 7 distinct URLs. But only 5 of those URLs are on learn.microsoft.com; source 6 is https://a2a-protocol.org/latest/specification/ and source 7 is https://www.jsonrpc.org/specification. The microsoft-learn MCP server serves learn.microsoft.com and nothing else, so 2 of the 7 cannot have come from MCP research, and the qualifier 'fra MCP-research' overclaims for them. This is the counting class's provenance sibling: the defect is not in the number but in what the number is asserted to be a count OF. The contrast inside the same batch makes the shape visible - idx-26n's line on the neighbouring file names 'MCP-research + tavily-research' and so does cover its non-Learn sources, while this one names MCP alone and does not. NOT REPAIRED: the ratified footer form governs counts and their referent, and it has never been extended to provenance qualifiers - doing so unasked would widen a form that stands in 21 files. Candidate resolutions, neither ratified: name the actual provenance mix, or drop the provenance qualifier and let the line be a pure count. Raised as an open operator question in STATE.\nRATIFIED 2026-08-11 - open question #21 is settled. The governing form is recorded ONCE, in _meta.ratified_forms[\"#21\"] of this queue; it is deliberately not restated here, because two copies of a policy drift apart. CONSEQUENCE HERE: delete the qualifier only - \"**Total sources cited**: 7 unique URLs fra MCP-research\" becomes \"**Total sources cited**: 7 unique URLs\". The count is verified true (7 numbered entries, 7 distinct URLs) and is untouched. NOT YET APPLIED: ratification and application are separate acts, and this entry stays open until its edit is made and verified.",
"evidence": "footer of this file (last line after the idx-26o deletion); Kilder section sources 1-7, of which 6 (a2a-protocol.org) and 7 (jsonrpc.org) are not learn.microsoft.com; idx-26o resolution; idx-26t (sibling finding, same batch)",
"anchors": [
"**Total sources cited**: 7 unique URLs fra MCP-research"
]
],
"resolution": "APPLIED 2026-08-12 under _meta.ratified_forms[\"#21\"]. The qualifier is deleted and the count kept: the line now reads \"**Total sources cited**: 7 unique URLs\". Nothing else in the file changed, and the count was not re-opened - it was already measured true (7 numbered entries, 7 distinct URLs). Verified by whole-line comparison: the pre-edit line occurs 0x, the post-edit line exactly 1x."
},
{
"id": "idx-26x",
"file": "skills/ms-ai-advisor/references/copilot-extensibility/enterprise-governance-copilot-deployment.md",
"class": "multi-locator",
"status": "open",
"status": "resolved",
"raised": "2026-08-11",
"summary": "Booked 2026-08-11 from the hand-classified footer class measured in docs sec. 9.16: 58 members in 58 files, dataset scripts/kb-eval/data/r11-footer-class-2026-08-11.json (tracked). Operator decision #17 that day split the class - the 42 line-form unflagged members take one class-wide record, and the 16 that carry a decision the ratified form does not settle get their own entry. This is one of those 16. Reason: BLOCK FORM, no flags. Bucket ikke_sjekkbar - the block states per-tool counts with NO total, so no internal cross-check exists. MEASURED at :919-:924: a \"**MCP-statistikk:**\" label followed by five bullets - 3 microsoft_docs_search calls, 2 microsoft_docs_fetch calls, 1 microsoft_code_sample_search call, then \"25+ unike Microsoft Learn URLs referert\" and \"15+ PowerShell code samples inkludert\". THE RATIFIED FORM (idx-26j, 2026-08-09): current provenance, one referent for the whole footer block, and the unverifiable MCP-calls line is DELETED rather than corrected to the sum. The ground is UNVERIFIABILITY, not arithmetic - the consistency bucket governs BOOKKEEPING, never whether something is deleted. WHAT IS UNDECIDED: the last two bullets are NOT call counts. The URL bullet is the KEEP class (a source count). The code-samples bullet is neither a call count nor a source-URL count - it is a content inventory, a third category the ratified form has never addressed. Deleting only the three call bullets leaves the \"**MCP-statistikk:**\" label standing over two bullets that are not call statistics, so the label's referent silently changes under the edit - the same referent-drift the ratified form exists to stop. Also worth recording for whoever adjudicates: both survivors are approximations (\"25+\", \"15+\") with no exact referent, so they are not checkable either, but they are outside this class. Candidate scopes, NEITHER ratified: (a) delete the three call bullets and re-label the heading, (b) delete the heading and the three call bullets and re-attach the two survivors. OPEN QUESTION #22 APPLIES: \"delete the line\" is undefined for block form. This entry is booked, not repaired, and no deletion should be applied to it until the block-form deletion scope is ratified.\nRATIFIED 2026-08-11 - open question #22 is settled. The governing form is recorded ONCE, in _meta.ratified_forms[\"#22\"] of this queue; it is deliberately not restated here, because two copies of a policy drift apart. Any sentence above saying no deletion should be applied until the block-form deletion scope is ratified is hereby LIFTED. CONSEQUENCE HERE: \"**MCP-statistikk:**\" names the generation, so the label goes with the three call bullets. The two survivors (\"25+ unike Microsoft Learn URLs referert\", \"15+ PowerShell code samples inkludert\") are promoted to the enclosing level. No rename. Their being approximations is noted in this entry and is outside this class. NOT YET APPLIED: ratification and application are separate acts, and this entry stays open until its edit is made and verified.",
"evidence": "skills/ms-ai-advisor/references/copilot-extensibility/enterprise-governance-copilot-deployment.md:919-924 (the MCP-statistikk block, file ends at :925); dataset r11-footer-class-2026-08-11.json (form=block, b=ikke_sjekkbar, flags=[]); idx-26j resolution (ratified form); docs sec. 9.16",
@ -432,39 +433,42 @@
"- 1 microsoft_code_sample_search call",
"- 25+ unike Microsoft Learn URLs referert",
"- 15+ PowerShell code samples inkludert"
]
],
"resolution": "APPLIED 2026-08-12 under #22. \"**MCP-statistikk:**\" and its three call bullets are deleted; the two artifact-class survivors are promoted.\nFIRST APPLICATION OF \"PROMOTE\" IN THE WHOLE QUEUE, and it needed an operational definition the form does not supply. Checked before deciding: none of the 21 resolved entries had a survivor, so all 40 prior edits were pure deletions and no precedent existed. PROMOTION IS DEFINED HERE AS REMOVAL OF THE LIST MARKER, so the line becomes a standalone paragraph at the level the label stood on. Leaving the survivors as bullets would have made them read as a fourth and fifth item of the Verified/Baseline/Estimat list immediately above - referent drift, which is the exact failure #22 exists to prevent. Removing \"- \" authors no words, which is what makes it admissible under NEVER RENAME.\nA SECOND DEFECT SURFACED ONLY AFTER THE EDIT, and only by reading the rendered shape: two promoted lines with no blank line between them are ONE markdown paragraph, silently fusing \"25+ unike Microsoft Learn URLs referert\" and \"15+ PowerShell code samples inkludert\" into a single run-on claim. A blank line was inserted. The whole-line invariant passed both before and after this fix - it cannot see coherence, only presence."
},
{
"id": "idx-26y",
"file": "skills/ms-ai-advisor/references/copilot-extensibility/microsoft-graph-api-copilot-integration.md",
"class": "replacement",
"status": "open",
"status": "resolved",
"raised": "2026-08-11",
"summary": "Booked 2026-08-11 from the hand-classified footer class measured in docs sec. 9.16: 58 members in 58 files, dataset scripts/kb-eval/data/r11-footer-class-2026-08-11.json (tracked). Operator decision #17 that day split the class - the 42 line-form unflagged members take one class-wide record, and the 16 that carry a decision the ratified form does not settle get their own entry. This is one of those 16. Reason: flag provenance_mix, and the ONLY member of all 58 in bucket \"tvetydig\". MEASURED at :544: \"**MCP calls:** 7 (3x docs_search, 2x docs_fetch, 1x code_sample_search, 1x ToolSearch)\". The verdict FLIPS between two defensible readings. Read as \"tool calls of any kind\", 3+2+1+1 = 7 and the line is consistent. Read as \"MCP calls\" - which is what the label says - the microsoft-learn components sum to 6, ToolSearch is not a microsoft-learn MCP tool, and the stated 7 is ONE TOO HIGH. THAT DIRECTION MATTERS BEYOND THIS FILE: sec. 9.16 measured 10 of 11 inconsistent members as stated < sum, and this file under the MCP-only reading is one of only two places in the whole class where the deviation runs the other way. It is part of the evidence that killed \"all deviations run the same way\" as an ABSOLUTE claim - the 91% tendency survives, the absolute does not. THE RATIFIED FORM (idx-26j, 2026-08-09): current provenance, one referent for the whole footer block, and the unverifiable MCP-calls line is DELETED rather than corrected to the sum. The ground is UNVERIFIABILITY, not arithmetic - the consistency bucket governs BOOKKEEPING, never whether something is deleted. OPEN QUESTION #21 APPLIES: this is the idx-26w provenance class - the defect is not the number but what the number is asserted to be a count OF. STATE records that all four members of this class (idx-26w plus the three booked here) are to be decided TOGETHER, not one at a time. Neighbour :543 \"**Total kilder:** 11 MCP-verified URLs, 4+ code samples\" is the KEEP class (a source count) and is anchored here so a deletion cannot take it silently.\nRATIFIED 2026-08-11 - open question #21 is settled. The governing form is recorded ONCE, in _meta.ratified_forms[\"#21\"] of this queue; it is deliberately not restated here, because two copies of a policy drift apart. CONSEQUENCE HERE: #21 does NOT apply to this entry. Measured - all 13 distinct URLs in this file are on learn.microsoft.com, so the keep-class neighbour \"**Total kilder:** 11 MCP-verified URLs, 4+ code samples\" carries no overclaim and is untouched. The anchored line \"**MCP calls:** 7 (...)\" is a call count and is deleted by the idx-26j form, which disposes of its ToolSearch provenance problem as a side effect rather than by a decision of its own. NOT YET APPLIED: ratification and application are separate acts, and this entry stays open until its edit is made and verified.",
"evidence": "skills/ms-ai-advisor/references/copilot-extensibility/microsoft-graph-api-copilot-integration.md:543-544 (file ends at :545); dataset r11-footer-class-2026-08-11.json (form=line, b=tvetydig, flags=[provenance_mix]); idx-26w (same provenance class, open); docs sec. 9.16 point 5 (the direction finding)",
"anchors": [
"**MCP calls:** 7 (3x docs_search, 2x docs_fetch, 1x code_sample_search, 1x ToolSearch)",
"**Total kilder:** 11 MCP-verified URLs, 4+ code samples"
]
],
"resolution": "APPLIED 2026-08-12 under the idx-26j form, NOT under #21, exactly as this entry's own ratification addendum records. The call count \"**MCP calls:** 7 (...)\" is deleted and its ToolSearch provenance problem dies with it; the keep-class neighbour \"**Total kilder:** 11 MCP-verified URLs, 4+ code samples\" is untouched, on the measurement that all 13 distinct URLs in the file are on learn.microsoft.com.\nBOOKED SEPARATELY AS idx-26as: this same file carries 18 \"**Confidence:** Verified (microsoft-learn docs_*)\" markers and 10 \"*Verified via microsoft_docs_*\" markers, all naming tools used. It is the densest instance of a class #22 does not govern, and it was left standing deliberately rather than swept in."
},
{
"id": "idx-26z",
"file": "skills/ms-ai-advisor/references/prompt-engineering/chain-of-thought-prompting.md",
"class": "multi-locator",
"status": "open",
"status": "resolved",
"raised": "2026-08-11",
"summary": "Booked 2026-08-11 from the hand-classified footer class measured in docs sec. 9.16: 58 members in 58 files, dataset scripts/kb-eval/data/r11-footer-class-2026-08-11.json (tracked). Operator decision #17 that day split the class - the 42 line-form unflagged members take one class-wide record, and the 16 that carry a decision the ratified form does not settle get their own entry. This is one of those 16. Reason: BLOCK FORM plus flag direction_stated_gt_enumerated. Bucket inkonsistent. MEASURED at :494-:500: a \"### MCP-kall utført\" heading, three numbered items (microsoft_docs_search, microsoft_code_sample_search, microsoft_docs_fetch), then \"**Totalt:** 4 MCP-kall, 3 unike Microsoft Learn-kilder.\" THIS IS THE SINGLE COUNTEREXAMPLE IN THE CLASS: stated 4 EXCEEDS the enumerated 3, while all ten other inconsistent members state a total BELOW their sum. sec. 9.16 records it as the finding that falsified \"all deviations run the same way\" as an absolute; the tendency (91%) survives, the absolute does not, and the earlier \"zero the other way\" was an artefact of measuring half the population. THE RATIFIED FORM (idx-26j, 2026-08-09): current provenance, one referent for the whole footer block, and the unverifiable MCP-calls line is DELETED rather than corrected to the sum. The ground is UNVERIFIABILITY, not arithmetic - the consistency bucket governs BOOKKEEPING, never whether something is deleted. WHAT IS UNDECIDED: the total line is MIXED. One sentence carries a call count (4 MCP-kall) AND a source count (3 unike Microsoft Learn-kilder). Deleting the line as written would delete a KEEP-class claim along with the call count. Splitting it means AUTHORING a replacement sentence - which is precisely the hazard named in idx-26v, \"the replacement authors the defect class it closes\". Note also that the surviving source count was NOT verified here; this entry books the counting defect, it does not certify that 3 unique Learn sources is true. OPEN QUESTION #22 APPLIES: \"delete the line\" is undefined for block form. This entry is booked, not repaired, and no deletion should be applied to it until the block-form deletion scope is ratified.\nRATIFIED 2026-08-11 - open questions #22 AND #23 are both settled, and this entry is governed by BOTH. The two forms are recorded ONCE each, in _meta.ratified_forms[\"#22\"] and _meta.ratified_forms[\"#23\"] of this queue; neither is restated here, because two copies of a policy drift apart. Any sentence above saying no deletion should be applied until the block-form deletion scope is ratified is hereby LIFTED. CONSEQUENCE HERE: both forms apply to one edit. \"### MCP-kall utfoert\" plus the three numbered call items go under #22; the mixed total goes whole under #23, because both of its numbers are false (3 stated against 5 rows in this file's own Verified table, 4 stated against 3 enumerated). Nothing survives the block, so nothing is promoted. NOT YET APPLIED: ratification and application are separate acts, and this entry stays open until its edit is made and verified.",
"evidence": "skills/ms-ai-advisor/references/prompt-engineering/chain-of-thought-prompting.md:494-500 (file ends at :505); dataset r11-footer-class-2026-08-11.json (form=block, b=inkonsistent, flags=[direction_stated_gt_enumerated]); docs sec. 9.16 point 5; idx-26v (replacement-authors-the-defect hazard)",
"anchors": [
"### MCP-kall utført",
"**Totalt:** 4 MCP-kall, 3 unike Microsoft Learn-kilder."
]
],
"resolution": "APPLIED 2026-08-12 as ONE edit under both #22 and #23. \"### MCP-kall utført\" plus the three numbered call items go under #22; \"**Totalt:** 4 MCP-kall, 3 unike Microsoft Learn-kilder.\" goes whole under #23. Nothing survived the block, so nothing was promoted. The file's own \"### Microsoft Learn (Verified via MCP)\" table still stands with 5 rows and 5 distinct Learn URLs, and still falsifies the \"3 unike\" that was deleted - deletion was a correction here, not a loss."
},
{
"id": "idx-26aa",
"file": "skills/ms-ai-advisor/references/prompt-engineering/domain-specific-prompt-optimization.md",
"class": "multi-locator",
"status": "open",
"status": "resolved",
"raised": "2026-08-11",
"summary": "Booked 2026-08-11 from the hand-classified footer class measured in docs sec. 9.16: 58 members in 58 files, dataset scripts/kb-eval/data/r11-footer-class-2026-08-11.json (tracked). Operator decision #17 that day split the class - the 42 line-form unflagged members take one class-wide record, and the 16 that carry a decision the ratified form does not settle get their own entry. This is one of those 16. Reason: BLOCK FORM, no flags. Bucket ikke_sjekkbar - no total, so nothing to cross-check against. MEASURED at :587-:591: a \"### Verifiseringsmetode\" heading and three bullets - \"**MCP-søk** — 3 søk mot microsoft-learn (2026-02-04)\", \"**Fetch** — 2 fullstendige dokumenter hentet via microsoft_docs_fetch\", and \"**Code samples** — Søk mot microsoft_code_sample_search (ingen direkte treff ...)\". THIS FILE IS PART OF THE EVIDENCE FOR THE FOURTH ESCAPE FORM (sec. 9.16 point 2): the count rides on the ACTION WORD (\"MCP-søk ... 3 søk\"), not on the word \"kall\" or a tool identifier, so sweep A was blind to it and :589 was rescued only by its neighbouring line. THE RATIFIED FORM (idx-26j, 2026-08-09): current provenance, one referent for the whole footer block, and the unverifiable MCP-calls line is DELETED rather than corrected to the sum. The ground is UNVERIFIABILITY, not arithmetic - the consistency bucket governs BOOKKEEPING, never whether something is deleted. WHAT IS UNDECIDED: the third bullet records a call WITH NO NUMBER AT ALL. \"Delete the line\" is undefined for a call record that carries no count - it is neither a false number nor a checkable one, and the ratified form was written for a stated count. Candidate scopes, none ratified: delete all three bullets and the heading; delete the two numbered bullets only; or treat the numberless bullet as outside the class. OPEN QUESTION #22 APPLIES: \"delete the line\" is undefined for block form. This entry is booked, not repaired, and no deletion should be applied to it until the block-form deletion scope is ratified.\nRATIFIED 2026-08-11 - open question #22 is settled. The governing form is recorded ONCE, in _meta.ratified_forms[\"#22\"] of this queue; it is deliberately not restated here, because two copies of a policy drift apart. Any sentence above saying no deletion should be applied until the block-form deletion scope is ratified is hereby LIFTED. CONSEQUENCE HERE: \"### Verifiseringsmetode\" names the generation method, so heading and all three bullets go. No keep-class line stands under it, so nothing is promoted. The neighbouring \"**Dato generert:** 2026-02-04\" is a generation timestamp and is delete class under the same test. NOT YET APPLIED: ratification and application are separate acts, and this entry stays open until its edit is made and verified.",
"evidence": "skills/ms-ai-advisor/references/prompt-engineering/domain-specific-prompt-optimization.md:587-591 (file ends at :603); dataset r11-footer-class-2026-08-11.json (form=block, b=ikke_sjekkbar, flags=[]); docs sec. 9.16 point 2 (fourth escape form, this file named)",
@ -472,7 +476,8 @@
"### Verifiseringsmetode",
"- **MCP-søk** — 3 søk mot microsoft-learn (2026-02-04)",
"- **Fetch** — 2 fullstendige dokumenter hentet via microsoft_docs_fetch"
]
],
"resolution": "APPLIED 2026-08-12 under #22. \"### Verifiseringsmetode\" and all three bullets are deleted - the third (\"- **Code samples** — Søk mot microsoft_code_sample_search ...\") is not an anchor but is named by this entry's consequence and went with the block. The neighbouring generation timestamp \"**Dato generert:** 2026-02-04\" in the file footer is deleted under the same referent test.\nLEFT STANDING, booked as idx-26as: the line directly below it, \"**Generert av:** Cosmo Skyberg (AI Architect) via MCP-research\", is generation-class by the same test and names both the generator and the process. Removing it belongs to the Cosmo track (R13/R14); doing it here would cross a ratified boundary to save one line."
},
{
"id": "idx-26ab",
@ -491,7 +496,7 @@
"id": "idx-26ac",
"file": "skills/ms-ai-engineering/references/mlops-genaiops/infrastructure-as-code-mlops.md",
"class": "multi-locator",
"status": "open",
"status": "resolved",
"raised": "2026-08-11",
"summary": "Booked 2026-08-11 from the hand-classified footer class measured in docs sec. 9.16: 58 members in 58 files, dataset scripts/kb-eval/data/r11-footer-class-2026-08-11.json (tracked). Operator decision #17 that day split the class - the 42 line-form unflagged members take one class-wide record, and the 16 that carry a decision the ratified form does not settle get their own entry. This is one of those 16. Reason: BLOCK FORM, no flags. Bucket ikke_sjekkbar - per-tool counts with no call total. MEASURED at :932-:937: a \"**MCP-research metadata:**\" label and five bullets - search 4, fetch 3, code_sample 1, then \"**Total sources:** 10\" and \"**Dato for research:** 2026-06-19\". THE RATIFIED FORM (idx-26j, 2026-08-09): current provenance, one referent for the whole footer block, and the unverifiable MCP-calls line is DELETED rather than corrected to the sum. The ground is UNVERIFIABILITY, not arithmetic - the consistency bucket governs BOOKKEEPING, never whether something is deleted. WHAT IS UNDECIDED: the block mixes three categories under one label. The three call bullets are the class. \"**Total sources:** 10\" is the KEEP class. \"**Dato for research:** 2026-06-19\" is a GENERATION TIMESTAMP - a fourth thing. Under the referent ratified on idx-26j (current provenance, one referent for the whole block) a generation-time date falls in the same category as a generation-time call count, which would make its deletion a ONE-STEP CONSEQUENCE OF AN EXISTING RATIFICATION rather than a new form - exactly the reasoning recorded for the file-size line in idx-26u. That consequence is NAMED HERE, NOT APPLIED: extending the form unasked is what idx-26w refused to do, and the same restraint applies. OPEN QUESTION #22 APPLIES: \"delete the line\" is undefined for block form. This entry is booked, not repaired, and no deletion should be applied to it until the block-form deletion scope is ratified.\nRATIFIED 2026-08-11 - open question #22 is settled. The governing form is recorded ONCE, in _meta.ratified_forms[\"#22\"] of this queue; it is deliberately not restated here, because two copies of a policy drift apart. Any sentence above saying no deletion should be applied until the block-form deletion scope is ratified is hereby LIFTED. CONSEQUENCE HERE: \"**MCP-research metadata:**\" names the generation, so label plus the three call bullets plus \"**Dato for research:** 2026-06-19\" all go. \"**Total sources:** 10\" is the sole survivor and is promoted. NOT YET APPLIED: ratification and application are separate acts, and this entry stays open until its edit is made and verified.",
"evidence": "skills/ms-ai-engineering/references/mlops-genaiops/infrastructure-as-code-mlops.md:932-937 (file ends at :949); dataset r11-footer-class-2026-08-11.json (form=block, b=ikke_sjekkbar, flags=[]); idx-26u (generation-time metadata, same category argument); idx-26j resolution",
@ -501,13 +506,14 @@
"- **microsoft_docs_fetch calls:** 3",
"- **microsoft_code_sample_search calls:** 1",
"- **Total sources:** 10"
]
],
"resolution": "APPLIED 2026-08-12 under #22. \"**MCP-research metadata:**\", its three call bullets and \"- **Dato for research:** 2026-06-19\" are deleted; \"**Total sources:** 10\" is promoted (list marker removed, per the definition set in idx-26x). It now stands between the numbered source list and \"**Confidence levels:**\" as a standalone fact, which is what it is."
},
{
"id": "idx-26ad",
"file": "skills/ms-ai-engineering/references/mlops-genaiops/model-versioning-registry-management.md",
"class": "multi-locator",
"status": "open",
"status": "resolved",
"raised": "2026-08-11",
"summary": "Booked 2026-08-11 from the hand-classified footer class measured in docs sec. 9.16: 58 members in 58 files, dataset scripts/kb-eval/data/r11-footer-class-2026-08-11.json (tracked). Operator decision #17 that day split the class - the 42 line-form unflagged members take one class-wide record, and the 16 that carry a decision the ratified form does not settle get their own entry. This is one of those 16. Reason: BLOCK FORM, no flags. Bucket ikke_sjekkbar - no call total. MEASURED at :579-:584: a \"### MCP research summary\" heading and five bullets - \"**Total searches:** 3\", \"**Document fetches:** 2\", \"**Code samples:** 1\", \"**Unique sources:** 7 Microsoft Learn articles\", \"**Research timestamp:** April 2026\". THIS FILE IS THE STRONGEST EVIDENCE FOR THE FOURTH ESCAPE FORM: sec. 9.16 point 2 records that it was invisible IN ITS ENTIRETY to sweeps A and B, because the counting rides on \"Total searches\" - no \"kall\", no tool identifier next to a digit - and it entered the class ONLY through sweep D. A single sweep cannot prove its own completeness; this file is what that costs when it is assumed. THE RATIFIED FORM (idx-26j, 2026-08-09): current provenance, one referent for the whole footer block, and the unverifiable MCP-calls line is DELETED rather than corrected to the sum. The ground is UNVERIFIABILITY, not arithmetic - the consistency bucket governs BOOKKEEPING, never whether something is deleted. WHAT IS UNDECIDED, AND THIS ONE IS SHARP: the KEEP-class line \"**Unique sources:** 7 Microsoft Learn articles\" sits in the SAME BULLET LIST as the three call lines. Any scope defined at list level - \"delete the MCP research summary block\" - takes a keep-class claim with it. The timestamp bullet raises the same category question as idx-26ac and idx-26u. OPEN QUESTION #22 APPLIES: \"delete the line\" is undefined for block form. This entry is booked, not repaired, and no deletion should be applied to it until the block-form deletion scope is ratified.\nRATIFIED 2026-08-11 - open question #22 is settled. The governing form is recorded ONCE, in _meta.ratified_forms[\"#22\"] of this queue; it is deliberately not restated here, because two copies of a policy drift apart. Any sentence above saying no deletion should be applied until the block-form deletion scope is ratified is hereby LIFTED. CONSEQUENCE HERE: \"### MCP research summary\" names the generation, so heading plus Total searches, Document fetches, Code samples and Research timestamp all go - the sharp part of this entry, a keep-class line sitting in the SAME bullet list, is resolved by promoting it rather than by scoping at list level. \"**Unique sources:** 7 Microsoft Learn articles\" survives and is promoted. NOT YET APPLIED: ratification and application are separate acts, and this entry stays open until its edit is made and verified.",
"evidence": "skills/ms-ai-engineering/references/mlops-genaiops/model-versioning-registry-management.md:579-584 (file ends at :590); dataset r11-footer-class-2026-08-11.json (form=block, b=ikke_sjekkbar, flags=[]); docs sec. 9.16 point 2 (this file named as invisible to sweeps A and B); idx-26ac and idx-26u (timestamp category)",
@ -517,7 +523,8 @@
"- **Document fetches:** 2 (Full registry guide, MLflow management guide)",
"- **Code samples:** 1 (MLflow Python SDK examples)",
"- **Unique sources:** 7 Microsoft Learn articles"
]
],
"resolution": "APPLIED 2026-08-12 under #22. \"### MCP research summary\", the three call bullets and \"- **Research timestamp:** April 2026\" are deleted; \"**Unique sources:** 7 Microsoft Learn articles\" is promoted. The sharp part of this entry - a keep-class line sitting in the same bullet list as delete-class ones - was resolved by promoting it, not by scoping at list level.\nRESIDUE FOUND BESIDE THE EDIT and left standing, booked as idx-26as: the heading at :529 reads \"### Microsoft Learn (Verified via MCP research, April 2026)\". It restates BOTH the timestamp and the provenance this edit had just removed, in the same file. The edit is correct; it is simply not complete for the file, and completing it needs a form this queue has not ratified."
},
{
"id": "idx-26ae",
@ -536,7 +543,7 @@
"id": "idx-26af",
"file": "skills/ms-ai-governance/references/responsible-ai/ai-act-annex-iii-checklist.md",
"class": "replacement",
"status": "open",
"status": "resolved",
"raised": "2026-08-11",
"summary": "Booked 2026-08-11 from the hand-classified footer class measured in docs sec. 9.16: 58 members in 58 files, dataset scripts/kb-eval/data/r11-footer-class-2026-08-11.json (tracked). Operator decision #17 that day split the class - the 42 line-form unflagged members take one class-wide record, and the 16 that carry a decision the ratified form does not settle get their own entry. This is one of those 16. Reason: BLOCK FORM plus flag provenance_mix. Bucket ikke_sjekkbar - no call total. MEASURED at :521-:525: the heading \"### MCP-søk utført 2026-02\" over three bullets - \"microsoft_docs_search: 2 queries\", \"WebSearch: 4 queries\", \"tavily_extract: 5 URLs\". TWO OF THE THREE TOOLS ARE NOT MCP microsoft-learn. The heading asserts MCP provenance for all three, so the overclaim is structural: it is carried by the heading, not by any single bullet, which is a shape idx-26w did not have (there the qualifier sat inside the counted line itself). A second, independent problem: the three bullets do not share a unit - \"queries\" twice and \"URLs\" once - so even summing them is not well defined, and no total is stated. THE RATIFIED FORM (idx-26j, 2026-08-09): current provenance, one referent for the whole footer block, and the unverifiable MCP-calls line is DELETED rather than corrected to the sum. The ground is UNVERIFIABILITY, not arithmetic - the consistency bucket governs BOOKKEEPING, never whether something is deleted. OPEN QUESTION #21 APPLIES: this is the idx-26w provenance class - the defect is not the number but what the number is asserted to be a count OF. STATE records that all four members of this class (idx-26w plus the three booked here) are to be decided TOGETHER, not one at a time. Neighbour \"**Total sources referenced:** 8 (5 primary, 3 secondary)\" is the KEEP class and is anchored so it cannot be removed with the block.\nRATIFIED 2026-08-11 - open question #22 is settled. The governing form is recorded ONCE, in _meta.ratified_forms[\"#22\"] of this queue; it is deliberately not restated here, because two copies of a policy drift apart. Any sentence above saying no deletion should be applied until the block-form deletion scope is ratified is hereby LIFTED. CONSEQUENCE HERE: this entry is governed by #22, NOT by #21. \"### MCP-soek utfoert 2026-02\" names the generation and carries the overclaim structurally, so heading and all three bullets go together and the overclaim dies with them - no provenance decision is needed. \"**Total sources referenced:** 8 (5 primary, 3 secondary)\" carries no qualifier, survives, and is promoted. NOT YET APPLIED: ratification and application are separate acts, and this entry stays open until its edit is made and verified.",
"evidence": "skills/ms-ai-governance/references/responsible-ai/ai-act-annex-iii-checklist.md:521-527 (file ends at :534); dataset r11-footer-class-2026-08-11.json (form=block, b=ikke_sjekkbar, flags=[provenance_mix]); idx-26w (same class, open); STATE open question #21",
@ -546,13 +553,14 @@
"- `WebSearch`: 4 queries (Annex III categories, classification criteria, forvaltningsloven, Art. 6(3))",
"- `tavily_extract`: 5 URLs (Official Annex III text, Article 6, DPO guide, Hjort analysis)",
"**Total sources referenced:** 8 (5 primary, 3 secondary)"
]
],
"resolution": "APPLIED 2026-08-12 under #22, NOT #21. The heading \"### MCP-søk utført 2026-02\" carried the overclaim structurally, so it died with its three bullets and no provenance decision was needed - the scope narrowing recorded in #21 held in practice. \"**Total sources referenced:** 8 (5 primary, 3 secondary)\" was already at the enclosing level, so it needed no promotion, only to move up."
},
{
"id": "idx-26ag",
"file": "skills/ms-ai-governance/references/responsible-ai/ai-act-compliance-guide.md",
"class": "replacement",
"status": "open",
"status": "resolved",
"raised": "2026-08-11",
"summary": "Booked 2026-08-11 from the hand-classified footer class measured in docs sec. 9.16: 58 members in 58 files, dataset scripts/kb-eval/data/r11-footer-class-2026-08-11.json (tracked). Operator decision #17 that day split the class - the 42 line-form unflagged members take one class-wide record, and the 16 that carry a decision the ratified form does not settle get their own entry. This is one of those 16. Reason: BLOCK FORM plus flag provenance_mix. Bucket ikke_sjekkbar - no call total. MEASURED at :732-:735: the label \"**MCP-søk utført 2026-02-03:**\" over three bullets - \"microsoft_docs_search: 3 queries\", \"WebSearch: 2 queries\", \"microsoft_docs_fetch: 3 URLs\". Same structural shape as idx-26af: the MCP claim is carried by the LABEL and one of the three tools (WebSearch) is not MCP microsoft-learn, so the overclaim cannot be repaired by editing any single bullet. Units again differ across bullets (queries, queries, URLs) with no stated total. THE RATIFIED FORM (idx-26j, 2026-08-09): current provenance, one referent for the whole footer block, and the unverifiable MCP-calls line is DELETED rather than corrected to the sum. The ground is UNVERIFIABILITY, not arithmetic - the consistency bucket governs BOOKKEEPING, never whether something is deleted. OPEN QUESTION #21 APPLIES: this is the idx-26w provenance class - the defect is not the number but what the number is asserted to be a count OF. STATE records that all four members of this class (idx-26w plus the three booked here) are to be decided TOGETHER, not one at a time. Neighbour \"**Total sources referenced:** 16 (7 primary, 9 secondary/tertiary)\" is the KEEP class and is anchored.\nRATIFIED 2026-08-11 - open question #22 is settled. The governing form is recorded ONCE, in _meta.ratified_forms[\"#22\"] of this queue; it is deliberately not restated here, because two copies of a policy drift apart. Any sentence above saying no deletion should be applied until the block-form deletion scope is ratified is hereby LIFTED. CONSEQUENCE HERE: this entry is governed by #22, NOT by #21. Same shape as idx-26af - \"**MCP-soek utfoert 2026-02-03:**\" names the generation and carries the overclaim, so label and all three bullets go and the overclaim dies with them. \"**Total sources referenced:** 16 (7 primary, 9 secondary/tertiary)\" survives and is promoted. NOT YET APPLIED: ratification and application are separate acts, and this entry stays open until its edit is made and verified.",
"evidence": "skills/ms-ai-governance/references/responsible-ai/ai-act-compliance-guide.md:732-737 (file ends at :744); dataset r11-footer-class-2026-08-11.json (form=block, b=ikke_sjekkbar, flags=[provenance_mix]); idx-26w and idx-26af (same class); STATE open question #21",
@ -562,7 +570,8 @@
"- `WebSearch`: 2 queries (EU AI Act 2026 requirements, Norway implementation)",
"- `microsoft_docs_fetch`: 3 URLs (Compliance Manager, AI governance guides)",
"**Total sources referenced:** 16 (7 primary, 9 secondary/tertiary)"
]
],
"resolution": "APPLIED 2026-08-12 under #22, NOT #21 - same shape as idx-26af. \"**MCP-søk utført 2026-02-03:**\" and its three bullets are deleted and the overclaim dies with them. \"**Total sources referenced:** 16 (7 primary, 9 secondary/tertiary)\" survives untouched at the enclosing level."
},
{
"id": "idx-26ah",
@ -580,7 +589,7 @@
"id": "idx-26ai",
"file": "skills/ms-ai-governance/references/responsible-ai/ai-risk-taxonomy-classification.md",
"class": "multi-locator",
"status": "open",
"status": "resolved",
"raised": "2026-08-11",
"summary": "Booked 2026-08-11 from the hand-classified footer class measured in docs sec. 9.16: 58 members in 58 files, dataset scripts/kb-eval/data/r11-footer-class-2026-08-11.json (tracked). Operator decision #17 that day split the class - the 42 line-form unflagged members take one class-wide record, and the 16 that carry a decision the ratified form does not settle get their own entry. This is one of those 16. Reason: BLOCK FORM, no flags. Bucket ikke_sjekkbar - per-tool counts with no call total. MEASURED at :458-:463: a \"### MCP Calls Summary\" heading over four bullets - search 3 calls, fetch 2 calls, code_sample 1 call, then \"**Totalt unike URLer:** 6 verified Microsoft Learn articles\". THE RATIFIED FORM (idx-26j, 2026-08-09): current provenance, one referent for the whole footer block, and the unverifiable MCP-calls line is DELETED rather than corrected to the sum. The ground is UNVERIFIABILITY, not arithmetic - the consistency bucket governs BOOKKEEPING, never whether something is deleted. WHAT IS UNDECIDED: the KEEP-class bullet sits inside the same list under a heading that names the whole block \"MCP Calls Summary\", so a heading-level scope removes a source count that is not part of the class - the idx-26ad problem in a second file. NOTED FOR THE ADJUDICATOR, NOT BOOKED: that surviving bullet has a label/value tension of its own - it counts ARTICLES under a URL label, the same shape as idx-26ae and idx-26ah - so deleting the call bullets does not leave a clean block behind. That is a keep-side observation and is deliberately left out of this entry's scope rather than widened into it. OPEN QUESTION #22 APPLIES: \"delete the line\" is undefined for block form. This entry is booked, not repaired, and no deletion should be applied to it until the block-form deletion scope is ratified.\nRATIFIED 2026-08-11 - open question #22 is settled. The governing form is recorded ONCE, in _meta.ratified_forms[\"#22\"] of this queue; it is deliberately not restated here, because two copies of a policy drift apart. Any sentence above saying no deletion should be applied until the block-form deletion scope is ratified is hereby LIFTED. CONSEQUENCE HERE: \"### MCP Calls Summary\" names the generation, so heading plus the three call bullets go. \"**Totalt unike URLer:** 6 verified Microsoft Learn articles\" survives and is promoted. NOT YET APPLIED: ratification and application are separate acts, and this entry stays open until its edit is made and verified.",
"evidence": "skills/ms-ai-governance/references/responsible-ai/ai-risk-taxonomy-classification.md:458-463 (file ends at :474); dataset r11-footer-class-2026-08-11.json (form=block, b=ikke_sjekkbar, flags=[]); idx-26ad (same list-level scope problem); idx-26ae, idx-26ah (label/value shape)",
@ -590,13 +599,14 @@
"- **microsoft_docs_fetch:** 2 calls (AI Risk Assessment, Code of Conduct)",
"- **microsoft_code_sample_search:** 1 call (AI risk assessment code examples)",
"- **Totalt unike URLer:** 6 verified Microsoft Learn articles"
]
],
"resolution": "APPLIED 2026-08-12 under #22. \"### MCP Calls Summary\" and its three call bullets are deleted; \"**Totalt unike URLer:** 6 verified Microsoft Learn articles\" is promoted.\nNOT TOUCHED, and named rather than swept: the \"### Sist verifisert\" section immediately below carries \"- **Metode:** MCP microsoft-learn server\", which is generation-class by #22's referent test but was neither anchored here nor named by this entry's consequence. Booked as idx-26as."
},
{
"id": "idx-26aj",
"file": "skills/ms-ai-governance/references/responsible-ai/algorithmic-accountability-auditability.md",
"class": "multi-locator",
"status": "open",
"status": "resolved",
"raised": "2026-08-11",
"summary": "Booked 2026-08-11 from the hand-classified footer class measured in docs sec. 9.16: 58 members in 58 files, dataset scripts/kb-eval/data/r11-footer-class-2026-08-11.json (tracked). Operator decision #17 that day split the class - the 42 line-form unflagged members take one class-wide record, and the 16 that carry a decision the ratified form does not settle get their own entry. This is one of those 16. Reason: BLOCK FORM, no flags. Bucket konsistent - MEASURED at :562-:565, \"### MCP Calls: 6\" followed by 3x microsoft_docs_search, 2x microsoft_docs_fetch, 1x microsoft_code_sample_search, and 3+2+1 = 6. Consistency does not save it: the ground is unverifiability. WHAT MAKES THIS ONE STRUCTURALLY DIFFERENT: THE TOTAL IS CARRIED BY A MARKDOWN HEADING, not by a body line. sec. 9.16 records the method finding that a digit-less heading can hide a whole structural form; this is its inverse - a heading that IS the count. \"Delete the line\" therefore means deleting a section heading, and the document structure around it is part of the edit: the block sits between \"---\" separators, and the very next heading \"### Unique Sources: 8 verified Microsoft Learn URLs\" is KEEP class at the SAME level and carries its own count in the same shape. Deleting one heading and keeping its structural twin is a judgement the ratified form does not make. THE RATIFIED FORM (idx-26j, 2026-08-09): current provenance, one referent for the whole footer block, and the unverifiable MCP-calls line is DELETED rather than corrected to the sum. The ground is UNVERIFIABILITY, not arithmetic - the consistency bucket governs BOOKKEEPING, never whether something is deleted. OPEN QUESTION #22 APPLIES: \"delete the line\" is undefined for block form. This entry is booked, not repaired, and no deletion should be applied to it until the block-form deletion scope is ratified.\nRATIFIED 2026-08-11 - open question #22 is settled. The governing form is recorded ONCE, in _meta.ratified_forms[\"#22\"] of this queue; it is deliberately not restated here, because two copies of a policy drift apart. Any sentence above saying no deletion should be applied until the block-form deletion scope is ratified is hereby LIFTED. CONSEQUENCE HERE: the heading IS the count, so \"### MCP Calls: 6\" is deleted together with its three bullets - the whole section, not a body line. Its structural twin \"### Unique Sources: 8 verified Microsoft Learn URLs\" is keep class and is untouched. The judgement this entry said the ratified form did not make is now made: the twin survives because its referent is the artifact, not because it looks alike. NOT YET APPLIED: ratification and application are separate acts, and this entry stays open until its edit is made and verified.",
"evidence": "skills/ms-ai-governance/references/responsible-ai/algorithmic-accountability-auditability.md:560-567 (file ends at :572); dataset r11-footer-class-2026-08-11.json (form=block, b=konsistent, flags=[]); docs sec. 9.16 (heading-borne count); idx-26ak (second heading-borne instance)",
@ -606,13 +616,14 @@
"- 2x `microsoft_docs_fetch`",
"- 1x `microsoft_code_sample_search`",
"### Unique Sources: 8 verified Microsoft Learn URLs"
]
],
"resolution": "APPLIED 2026-08-12 under #22. The heading IS the count, so \"### MCP Calls: 6\" was deleted as a whole section together with its three bullets - not as a body line. Its structural twin \"### Unique Sources: 8 verified Microsoft Learn URLs\" is untouched: it survives because its referent is the artifact, not because it looks alike."
},
{
"id": "idx-26ak",
"file": "skills/ms-ai-governance/references/responsible-ai/model-monitoring-drift-detection.md",
"class": "multi-locator",
"status": "open",
"status": "resolved",
"raised": "2026-08-11",
"summary": "Booked 2026-08-11 from the hand-classified footer class measured in docs sec. 9.16: 58 members in 58 files, dataset scripts/kb-eval/data/r11-footer-class-2026-08-11.json (tracked). Operator decision #17 that day split the class - the 42 line-form unflagged members take one class-wide record, and the 16 that carry a decision the ratified form does not settle get their own entry. This is one of those 16. Reason: BLOCK FORM, no flags. Bucket inkonsistent - MEASURED at :774-:777: \"### Total MCP Calls: 4\" followed by microsoft_docs_search 3 queries, microsoft_docs_fetch 2 deep reads, microsoft_code_sample_search 1 query. The components sum to 6 against a stated 4 - the class's dominant direction (stated < sum), which sec. 9.16 measured at 10 of 11. THE RATIFIED FORM (idx-26j, 2026-08-09): current provenance, one referent for the whole footer block, and the unverifiable MCP-calls line is DELETED rather than corrected to the sum. The ground is UNVERIFIABILITY, not arithmetic - the consistency bucket governs BOOKKEEPING, never whether something is deleted. STRUCTURALLY this is the second heading-borne total, the same shape as idx-26aj: the number lives in a \"###\" heading, so the edit is a structural one, and the following heading \"### Total Unique URLs: 9\" is KEEP class at the same level with its own two-bullet breakdown (Primary 4 + Secondary 5 = 9, internally consistent). Decide the two heading-borne cases together or the corpus ends up with two different treatments of one shape. OPEN QUESTION #22 APPLIES: \"delete the line\" is undefined for block form. This entry is booked, not repaired, and no deletion should be applied to it until the block-form deletion scope is ratified.\nRATIFIED 2026-08-11 - open question #22 is settled. The governing form is recorded ONCE, in _meta.ratified_forms[\"#22\"] of this queue; it is deliberately not restated here, because two copies of a policy drift apart. Any sentence above saying no deletion should be applied until the block-form deletion scope is ratified is hereby LIFTED. CONSEQUENCE HERE: identical treatment to idx-26aj, which is why the two were decided together. \"### Total MCP Calls: 4\" and its three bullets go; \"### Total Unique URLs: 9\" and its internally consistent 4+5 breakdown are untouched. NOT YET APPLIED: ratification and application are separate acts, and this entry stays open until its edit is made and verified.",
"evidence": "skills/ms-ai-governance/references/responsible-ai/model-monitoring-drift-detection.md:774-781 (file ends at :782); dataset r11-footer-class-2026-08-11.json (form=block, b=inkonsistent, flags=[]); idx-26aj (same heading-borne shape); docs sec. 9.16 point 5",
@ -622,26 +633,28 @@
"- microsoft_docs_fetch: 2 deep reads",
"- microsoft_code_sample_search: 1 query",
"### Total Unique URLs: 9"
]
],
"resolution": "APPLIED 2026-08-12 under #22, identical treatment to idx-26aj, which is why the two were decided together. \"### Total MCP Calls: 4\" and its three bullets are deleted; \"### Total Unique URLs: 9\" and its internally consistent 4+5 breakdown are untouched."
},
{
"id": "idx-26al",
"file": "skills/ms-ai-governance/references/responsible-ai/responsible-ai-framework-overview.md",
"class": "multi-locator",
"status": "open",
"status": "resolved",
"raised": "2026-08-11",
"summary": "Booked 2026-08-11 from the hand-classified footer class measured in docs sec. 9.16: 58 members in 58 files, dataset scripts/kb-eval/data/r11-footer-class-2026-08-11.json (tracked). Operator decision #17 that day split the class - the 42 line-form unflagged members take one class-wide record, and the 16 that carry a decision the ratified form does not settle get their own entry. This is one of those 16. Reason: BLOCK FORM (two lines), no flags. Bucket ikke_sjekkbar - no total. MEASURED at :374-:375: \"**MCP-søk utført:** 3 søk (microsoft-learn)\" and \"**Sider hentet:** 3 full-fetch (concept-responsible-ai, assurance-ai, concept-responsible-ai-dashboard)\". Another fourth-escape-form instance (sec. 9.16 point 2): the count rides on the action word \"MCP-søk utført\", with no \"kall\" and no tool identifier next to a digit. THE RATIFIED FORM (idx-26j, 2026-08-09): current provenance, one referent for the whole footer block, and the unverifiable MCP-calls line is DELETED rather than corrected to the sum. The ground is UNVERIFIABILITY, not arithmetic - the consistency bucket governs BOOKKEEPING, never whether something is deleted. WHAT IS UNDECIDED: the block is two lines with no heading and no wrapper, so there is no structural boundary to delete to - \"the line\" is ambiguous between one line and both. The second line differs from every other member of the class in one respect worth recording: IT NAMES THE THREE FETCHED DOCUMENTS, so unlike a bare count it is in principle checkable against the file's own source list. THAT CHECK WAS NOT RUN - it is a different check from the class's internal cross-check, and asserting a verdict without running it is the exact failure this programme keeps catching. Whoever adjudicates should either run it or record that it was not run. OPEN QUESTION #22 APPLIES: \"delete the line\" is undefined for block form. This entry is booked, not repaired, and no deletion should be applied to it until the block-form deletion scope is ratified.\nRATIFIED 2026-08-11 - open question #22 is settled. The governing form is recorded ONCE, in _meta.ratified_forms[\"#22\"] of this queue; it is deliberately not restated here, because two copies of a policy drift apart. Any sentence above saying no deletion should be applied until the block-form deletion scope is ratified is hereby LIFTED. CONSEQUENCE HERE: both anchored lines are call counts and both go. No keep-class line stands with them, so nothing is promoted. NOT YET APPLIED: ratification and application are separate acts, and this entry stays open until its edit is made and verified.",
"evidence": "skills/ms-ai-governance/references/responsible-ai/responsible-ai-framework-overview.md:373-375 (file ends at :381); dataset r11-footer-class-2026-08-11.json (form=block, b=ikke_sjekkbar, flags=[]); docs sec. 9.16 point 2",
"anchors": [
"**MCP-søk utført:** 3 søk (microsoft-learn)",
"**Sider hentet:** 3 full-fetch (concept-responsible-ai, assurance-ai, concept-responsible-ai-dashboard)"
]
],
"resolution": "APPLIED 2026-08-12 under #22. Both anchored lines are call counts and both are deleted; no keep-class line stood with them, so nothing was promoted.\n\"**Sist verifisert:** 2026-02-03\" directly above is left standing because this entry's consequence did not name it. Whether a bare verification date is generation-class under the idx-26j referent test is genuinely undecided, and it is part of what idx-26as puts to the operator rather than something to settle by reflex mid-edit."
},
{
"id": "idx-26am",
"file": "skills/ms-ai-security/references/ai-security-engineering/prompt-injection-defense-patterns.md",
"class": "multi-locator",
"status": "open",
"status": "resolved",
"raised": "2026-08-11",
"summary": "Booked 2026-08-11 from the hand-classified footer class measured in docs sec. 9.16: 58 members in 58 files, dataset scripts/kb-eval/data/r11-footer-class-2026-08-11.json (tracked). Operator decision #17 that day split the class - the 42 line-form unflagged members take one class-wide record, and the 16 that carry a decision the ratified form does not settle get their own entry. This is one of those 16. Reason: BLOCK FORM, no flags. Bucket ikke_sjekkbar - no total. MEASURED at :476-:480: a \"**Research Coverage:**\" label over four bullets - \"3 MCP microsoft-learn docs_search calls\", \"3 MCP microsoft-learn docs_fetch calls (full documentation)\", \"9 unique source URLs from Microsoft Learn\", and a \"Coverage:\" bullet that is prose with no number. THE RATIFIED FORM (idx-26j, 2026-08-09): current provenance, one referent for the whole footer block, and the unverifiable MCP-calls line is DELETED rather than corrected to the sum. The ground is UNVERIFIABILITY, not arithmetic - the consistency bucket governs BOOKKEEPING, never whether something is deleted. WHAT IS UNDECIDED: the label \"Research Coverage\" spans four different kinds of statement - two call counts (the class), one source count (KEEP), and one prose enumeration of topics. There is no scope that removes the class without either leaving a label over a changed referent or taking non-class content with it. This is the idx-26x problem in a second file, and the two should be decided under one rule rather than file by file. OPEN QUESTION #22 APPLIES: \"delete the line\" is undefined for block form. This entry is booked, not repaired, and no deletion should be applied to it until the block-form deletion scope is ratified.\nRATIFIED 2026-08-11 - open question #22 is settled. The governing form is recorded ONCE, in _meta.ratified_forms[\"#22\"] of this queue; it is deliberately not restated here, because two copies of a policy drift apart. Any sentence above saying no deletion should be applied until the block-form deletion scope is ratified is hereby LIFTED. CONSEQUENCE HERE: the label is KEPT - \"**Research Coverage:**\" names coverage, not process, and is the only such label among the 14. The two call bullets go; the source-URL bullet and the prose Coverage bullet stay under the unchanged label. No promotion, no rename, no referent drift. NOT YET APPLIED: ratification and application are separate acts, and this entry stays open until its edit is made and verified.",
"evidence": "skills/ms-ai-security/references/ai-security-engineering/prompt-injection-defense-patterns.md:476-480 (file ends at :484); dataset r11-footer-class-2026-08-11.json (form=block, b=ikke_sjekkbar, flags=[]); idx-26x (same mixed-label shape)",
@ -651,7 +664,8 @@
"- 3 MCP microsoft-learn docs_fetch calls (full documentation)",
"- 9 unique source URLs from Microsoft Learn",
"- Coverage: Prompt Shields, Security Benchmark (AI-2, AI-3), LLM Security Planning, Content Filtering"
]
],
"resolution": "APPLIED 2026-08-12 under #22. THE ONLY LABEL KEPT among the 14: \"**Research Coverage:**\" names coverage rather than process. The two call bullets are deleted; the source-URL bullet and the prose Coverage bullet remain as bullets under the unchanged label. No promotion (the label still encloses them), no rename, no referent drift."
},
{
"id": "idx-26an",
@ -829,7 +843,7 @@
"id": "idx-26ao",
"file": "skills/ms-ai-engineering/references/mlops-genaiops/genaiops-llm-specific-practices.md",
"class": "multi-locator",
"status": "open",
"status": "resolved",
"raised": "2026-08-11",
"summary": "PULLED OUT OF THE CLASS ENTRY idx-26an the same session it was written. It was classified 2026-08-11 as line-form with no flag and no open decision, and a residue screen run before any corpus edit falsified that. MEASURED at :383: \"**Totalt:** 18 kilder, 8 MCP-kall.\" THE LINE IS MIXED. 8 MCP-kall is the delete class - an unverifiable call count. 18 kilder is the KEEP class, and unlike the call count it is CHECKABLE: the file carries a heading \"### Microsoft Learn-kilder (18 dokumenter)\" at :356 over a numbered source list. So \"delete the line\", the ratified form, would delete a verifiable claim along with an unverifiable one. This is the same structural defect as idx-26z, where \"**Totalt:** 4 MCP-kall, 3 unike Microsoft Learn-kilder.\" mixes the two classes in one sentence - and the same hazard idx-26v named: splitting the line means AUTHORING a replacement, which is how a repair authors the defect class it closes. THE RATIFIED FORM (idx-26j, 2026-08-09) still governs the call count; what is undecided is the scope of the edit on a mixed line. Note also :6, \"**Konfidensgrad:** Høy (basert på 18 MCP-kilder fra Microsoft Learn)\" - the header restates the 18. That is the KEEP-class number, so it is not a residue of the call count, but any re-authoring of :383 has to leave :6 coherent with it. Decide together with idx-26z.\nADDENDUM 2026-08-11, later the same day - a dated correction, not a silent reopening (sec. 9.2). THE FORM CLASSIFICATION IS WRONG AND THE ANCHOR SET WAS INCOMPLETE. The dataset records this file as form=line. It is BLOCK FORM. MEASURED at :377-383: the heading \"### MCP-kall utfoert\" (spelled with the o-slash in the file) stands over three tool bullets - docs_search 3 soek, docs_fetch 3 hentinger, code_sample_search 2 soek - which sum to 8 and agree with the stated total. That is the same shape as idx-26z (chain-of-thought-prompting.md), which the SAME dataset classified as block. Identical structure, opposite classification.\nCONSEQUENCE, WHICH IS WHY THIS IS BOOKED RATHER THAN NOTED: applying the ratified \"delete the line\" to :383 alone would leave the heading standing over three call bullets asserting the same unverifiable claim - exactly the idx-26aq residue, this time produced by an edit whose own entry failed to anchor the survivors. The four locators are anchored here rather than described in prose, so none survives only in a summary (the 795d494 failure). THIS ENTRY THEREFORE FALLS UNDER OPEN QUESTION #22 AS WELL AS #23, and #22 must be ratified before any deletion is applied to it.\nHOW IT WAS FOUND, and what the net cost: a sweep of ALL 46 line-form members of the 58 for an enumeration hidden above the anchor (a heading naming MCP/calls/research, or two or more tool-naming list items, within 12 lines above). Three hits, hand-read. Two are net noise - inferencing-optimization-caching.md, where the matched heading is a code-samples section and the anchor already carries its own full breakdown on one line, and ai-impact-assessment-framework.md, where the matched lines carry no per-tool counts. This one is real. One of 46, so the misclassification is isolated, not systemic - measured rather than assumed.\nTHE KEEP-CLASS HALF IS MEASURED, NOT ASSUMED, and it settles the #23 question for this file without authoring anything: \"18 kilder\" is TRUE - the file enumerates 18 numbered sources - and it is restated TWICE more, at :6 in the header block and at :356 in the heading \"### Microsoft Learn-kilder (18 dokumenter)\". Deleting the whole mixed line therefore loses no checkable claim. Contrast idx-26z, where the same measurement falsifies the keep half instead: \"3 unike Microsoft Learn-kilder\" against 5 rows in that file's own Verified table.\nRATIFIED 2026-08-11 - open questions #22 AND #23 are both settled, and this entry is governed by BOTH. The two forms are recorded ONCE each, in _meta.ratified_forms[\"#22\"] and _meta.ratified_forms[\"#23\"] of this queue; neither is restated here, because two copies of a policy drift apart. Any sentence above saying no deletion should be applied until the block-form deletion scope is ratified is hereby LIFTED. CONSEQUENCE HERE: both forms apply to ONE edit, and splitting them across two would reproduce the idx-26aq residue this entry exists to record. \"### MCP-kall utfoert\" plus the three tool bullets go under #22; \"**Totalt:** 18 kilder, 8 MCP-kall.\" goes whole under #23, because the 18 is true and restated at :6 and :356. Nothing is promoted; the header at :6 is left exactly as it is, since 18 is keep class there too. NOT YET APPLIED: ratification and application are separate acts, and this entry stays open until its edit is made and verified.",
"evidence": "skills/ms-ai-engineering/references/mlops-genaiops/genaiops-llm-specific-practices.md:383 (the mixed total), :356 (the 18-source heading and numbered list), :6 (header restating 18); dataset r11-footer-class-2026-08-11.json (classified form=line, flags=[] - the gap this entry records); idx-26z (same mix), idx-26v (replacement-authors-the-defect hazard)",
@ -841,26 +855,28 @@
"- **microsoft_docs_search**: 3 søk (GenAIOps overview, LLMOps best practices, lifecycle)",
"- **microsoft_docs_fetch**: 3 hentinger (maturity model, genaiops-for-mlops, databricks llmops)",
"- **microsoft_code_sample_search**: 2 søk (evaluation Python code, monitoring code)"
]
],
"resolution": "APPLIED 2026-08-12 as ONE edit under both #22 and #23, exactly as this entry required - splitting it across two edits would have reproduced the idx-26aq residue this entry exists to record. \"### MCP-kall utført\" plus the three tool bullets go under #22; \"**Totalt:** 18 kilder, 8 MCP-kall.\" goes whole under #23. Nothing is promoted. The 18 survives twice over, at :6 and in the \"### Microsoft Learn-kilder (18 dokumenter)\" heading, both untouched - which is why deleting the mixed line lost nothing.\nRESIDUE LEFT STANDING, booked as idx-26as: \"**Verifiseringsdato:** 2026-02-04\" now sits immediately after the deleted block and is a generation timestamp by the same test that deleted the block."
},
{
"id": "idx-26ap",
"file": "skills/ms-ai-engineering/references/mlops-genaiops/mlops-security-access-control.md",
"class": "multi-locator",
"status": "open",
"status": "resolved",
"raised": "2026-08-11",
"summary": "PULLED OUT OF THE CLASS ENTRY idx-26an the same session it was written. It was classified 2026-08-11 as line-form with no flag and no open decision, and a residue screen run before any corpus edit falsified that. MEASURED: the footer at :744 states \"**MCP Calls:** 8 (microsoft-learn docs search + fetch, code samples)\", and the DOCUMENT HEADER at :5 restates the same unverifiable count - \"**Confidence:** HIGH — Basert på offisiell Microsoft Learn dokumentasjon (8 MCP-oppslag, 16 kilder)\". Deleting the footer line under the ratified form would leave the number alive in the file's own header block. That is the residue class G7 exists for, and it is the idx-26ab situation in a second file - except that here the survivor sits in the reference contract's bold-label header rather than in body prose, so the second edit touches the file's structured metadata and cannot be a bare line deletion. The header line also carries \"16 kilder\", a source count and therefore KEEP class, in the same sentence - so the second locator is itself mixed and needs re-authoring rather than removal. THE RATIFIED FORM (idx-26j, 2026-08-09) governs the call count in both places; what is undecided is how the header line is re-authored without asserting something new. NOT REPAIRED HERE. The 2026-08-11 hand classification applied prose_repeat to one file only and missed this one; the gap is recorded rather than quietly patched, and the header block of all 58 members was then swept for the same shape - it returned this file and genaiops-llm-specific-practices.md and nothing else.\nRATIFIED 2026-08-11 - open question #24 is settled. The governing form is recorded ONCE, in _meta.ratified_forms[\"#24\"] of this queue; it is deliberately not restated here, because two copies of a policy drift apart. CONSEQUENCE HERE: both locators go in one edit. The footer \"**MCP Calls:** 8 (...)\" at :744 is deleted by the idx-26j form; the header parenthetical \"(8 MCP-oppslag, 16 kilder)\" at :5 is deleted whole, because 16 is false against 8 enumerated sources and 8 MCP-oppslag is unverifiable. The header line keeps its field name and its remaining claim. NOT YET APPLIED: ratification and application are separate acts, and this entry stays open until its edit is made and verified.",
"evidence": "skills/ms-ai-engineering/references/mlops-genaiops/mlops-security-access-control.md:744 (footer count) and :5 (header restating 8 MCP-oppslag alongside 16 kilder); dataset r11-footer-class-2026-08-11.json (classified form=line, flags=[] - the gap this entry records); idx-26ab (prose-repeat precedent); memory: KB reference contract uses a bold-label header",
"anchors": [
"**MCP Calls:** 8 (microsoft-learn docs search + fetch, code samples)",
"**Confidence:** HIGH — Basert på offisiell Microsoft Learn dokumentasjon (8 MCP-oppslag, 16 kilder)"
]
],
"resolution": "APPLIED 2026-08-12 under #24, both locators in one edit. The footer \"**MCP Calls:** 8 (...)\" at :744 is deleted by the idx-26j form; the header parenthetical \"(8 MCP-oppslag, 16 kilder)\" at :5 is deleted whole, leaving \"**Confidence:** HIGH — Basert på offisiell Microsoft Learn dokumentasjon\". Zero words authored: the field name and the remaining claim are byte-identical to before. The structured reference header keeps its shape, which is what made the re-authoring hazard worth measuring away."
},
{
"id": "idx-26aq",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-cost-optimization.md",
"class": "multi-locator",
"status": "open",
"status": "resolved",
"raised": "2026-08-11",
"summary": "SURFACED BY APPLYING idx-26an, in the per-file residue search AFTER the edit rather than before it. Deleting the footer line \"**Total MCP calls:** 6\" at :396 was correct under the ratified form and it removed a contradiction, but it left the same unverifiable claim standing ten lines above, in a form no line-based screen enumerates: seven MCP calls listed one per line at :388-394, under a heading that carries the count itself.\nTWO DEFECTS, NOT ONE, and they need separate decisions.\n(a) THE ENUMERATION IS THE SAME UNVERIFIABLE CLASS AS THE DELETED TOTAL. The ratified ground for deletion is unverifiability - no artifact in the repo records MCP calls per file - and that ground applies with MORE force to a list asserting seven specific calls with specific query strings than to a bare number. The deletion did not narrow the claim; it removed the summary of it and left the detail.\n(b) THE HEADING REFERENT IS MISLABELLED, and the two readings disagree. Measured: the source block directly above is numbered 1..10 with no gaps, and \"**Unique URLs:** 10\" at :396 is therefore enumerable and TRUE - keep class, untouched. So \"(7 unique sources)\" is FALSE read as sources, since there are 10, and true only read as calls, since there are 7 bullets. Rewriting the parenthetical means authoring a new claim, which is the idx-26v hazard, so it is not a free rename.\nHEADING-BORNE, SO IT JOINS OPEN QUESTION #22 beside idx-26aj and idx-26ak rather than this file line-form. Same shape as idx-26ak: heading carrying the total, per-tool bullets beneath, a sibling keep-class count alongside. Deciding this one alone would decide #22 by precedent for the other two, so decide the three together.\nCORRECTION TO A RESOLVED ENTRY, written into idx-26j as a dated addendum in the same pass rather than left standing: idx-26j put this file :396 in its \"not checkable\" bucket, defined as a total with no per-tool numbers to sum. That is falsified - the per-tool numbers exist, they are just ten lines up and past a heading. The arithmetic test WAS applicable here, and it fails: 6 stated against 7 enumerated. The other three in that bucket were re-checked against HEAD before this was written and all three hold (multi-turn-conversation-management.md:683 and ai-services-networking-security.md:620 name tools without numbers; adaptive-cards-copilot-responses.md:517 is the mirror case, per-tool numbers with no total). One of four wrong, not four of four - the bucket is dented, not void.\nPOPULATION MEASURED BEFORE BOOKING: 5 files of 389 carry an enumerated MCP-call list; 1 is a false positive, 3 are already booked (idx-26z, idx-26ai, idx-26ak), and this is the only unbooked one. The form is small and now fully enumerated, not an open-ended sweep.\nRATIFIED 2026-08-11 - open question #22 is settled. The governing form is recorded ONCE, in _meta.ratified_forms[\"#22\"] of this queue; it is deliberately not restated here, because two copies of a policy drift apart. Any sentence above saying no deletion should be applied until the block-form deletion scope is ratified is hereby LIFTED. CONSEQUENCE HERE: \"### MCP-soek (7 unique sources)\" is heading-borne and names the generation, so it is deleted together with the seven enumerated calls at :388-394. The rename question raised in defect (b) therefore never has to be answered - the mislabelled parenthetical dies with the heading rather than being rewritten. \"**Unique URLs:** 10\" is keep class, verified enumerable against a 1..10 source block, and is untouched. NOT YET APPLIED: ratification and application are separate acts, and this entry stays open until its edit is made and verified.",
"evidence": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-cost-optimization.md:386-396 (file ends at :397); numbered source block at :331-384 enumerates 1..10; idx-26an resolution (residue sweep, 2026-08-11); idx-26aj and idx-26ak (same heading-borne shape, open question #22); idx-26j resolution and its 2026-08-11 addendum (the falsified not-checkable classification); dataset r11-footer-class-2026-08-11.json",
@ -870,7 +886,8 @@
"- microsoft_docs_fetch: `/azure/ai-services/commitment-tier`",
"- microsoft_docs_search: \"Azure OpenAI provisioned throughput PTU cost optimization\"",
"**Unique URLs:** 10"
]
],
"resolution": "APPLIED 2026-08-12 under #22. \"### MCP-søk (7 unique sources)\" is deleted together with all seven enumerated calls, so the mislabelled parenthetical died with the heading and the rename question raised in defect (b) never had to be answered. \"**Unique URLs:** 10\" is untouched and remains enumerable against the 1..10 source block directly above it."
},
{
"id": "idx-26ar",
@ -888,6 +905,38 @@
"- `microsoft_docs_fetch`: Mirroring overview, PostgreSQL architecture",
"- `microsoft_code_sample_search`: \"Fabric mirroring OneLake Delta Lake\""
]
},
{
"id": "idx-26as",
"file": "skills/ms-ai-advisor/references/copilot-extensibility/microsoft-graph-api-copilot-integration.md",
"class": "multi-locator",
"status": "open",
"raised": "2026-08-12",
"summary": "Booked while applying the 17 ratified entries, from the residue sweep run BESIDE those edits rather than over them - the twelfth-time discipline, which paid again.\nTHE FINDING: the delete-class referent named by _meta.ratified_forms[\"#22\"] - tools used, calls made, when the document was generated - also occurs OUTSIDE the footer blocks the 58-member class was scoped over, as PER-CLAIM provenance markers scattered through document bodies. #22 does not govern them, and this is a scope fact rather than a judgement on merit: #22 is a form about what \"delete the line\" means for a BLOCK and its LABEL, and all 14 of its members are footer blocks. Extending it silently would widen a ratified form far past what was measured - the hazard idx-26w recorded and the ground idx-26ar was left open on.\nMEASURED CORPUS-WIDE 2026-08-12 over skills/, four sub-forms:\n (A) 18 lines \"**Confidence:** Verified (microsoft-learn docs_fetch|docs_search)\", ALL in microsoft-graph-api-copilot-integration.md - the file idx-26y had just edited.\n (B) 11 lines \"*Verified via microsoft_docs_*\" - 10 in that same file, 1 as a heading qualifier at digdir-principle-1-user-centric-design.md:280.\n (C) 13 lines of inline prose provenance, of which 5 are the IDENTICAL heading \"### Verified (MCP-research)\" across five rag-architecture files.\n (D) 1 surviving generation timestamp, \"**Verifiseringsdato:** 2026-02-04\" at genaiops-llm-specific-practices.md:377, left standing immediately after the block idx-26ao deleted.\nD WAS COUNTED BEFORE IT WAS CLASSIFIED, which is the check that kept it at 1 instead of 2: the second grep hit, \"**Verifiseringsdato:** YYYY-MM-DD\" at regional-availability-verification.md:67, is a TEMPLATE PLACEHOLDER and not an instance of the class at all.\nCONCENTRATION IS THE POINT AND CHANGES WHAT A DECISION COSTS: 28 of the ~42 markers sit in ONE file. This is not evenly spread corpus rot, and a decision scoped to that single file would dispose of two thirds of the class.\nALSO IN THIS CLASS, named where they were found rather than folded in silently: \"- **Metode:** MCP microsoft-learn server\" (ai-risk-taxonomy-classification.md:463, under \"### Sist verifisert\"); \"### Microsoft Learn (Verified via MCP research, April 2026)\" (model-versioning-registry-management.md:529, which restates BOTH the timestamp and the provenance idx-26ad had just deleted, in the file that edit had just cleaned); and \"**Generert av:** Cosmo Skyberg (AI Architect) via MCP-research\" (domain-specific-prompt-optimization.md), whose removal belongs to the Cosmo track R13/R14 and not to this class.\nOPEN QUESTION FOR THE OPERATOR, deliberately NOT inherited: does the idx-26j referent test extend from footer blocks to per-claim markers - and if it does, is a bare \"**Confidence:** Verified\" the replacement, or does the whole marker go?\nWHY THIS IS A QUESTION AND NOT AN INHERITANCE, stated as the asymmetry rather than as caution: (A) and (B) differ from every member of the footer class in one way that may matter. They attach to a SPECIFIC claim rather than summarising the document, so deleting them removes a per-claim confidence signal and not merely an unverifiable provenance qualifier. The footer class had no such signal to lose. Whether that signal is worth keeping when its provenance half is unverifiable is exactly the trade the operator has not yet been asked to make.\nALSO UNSETTLED, and surfaced by idx-26al rather than by this sweep: whether a bare verification date (\"**Sist verifisert:** 2026-02-03\") is generation-class at all.\nNOT REPAIRED.",
"evidence": "corpus-wide grep over skills/ 2026-08-12: 18 (A) + 11 (B) + 13 (C) + 1 (D), placeholder at regional-availability-verification.md:67 excluded by inspection; residue sweep over the 17 files edited this session; _meta.ratified_forms[\"#22\"] scope paragraph (\"EXPLICITLY NOT COVERED\"); idx-26ar as the precedent for leaving an out-of-scope form explicitly open; idx-26w on the cost of widening a form unasked",
"anchors": [
"**Confidence:** Verified (microsoft-learn docs_fetch)",
{
"file": "skills/ms-ai-governance/references/norwegian-public-sector-governance/digdir-principle-1-user-centric-design.md",
"text": "### Microsoft-kilder for AI og tilgjengelighet (Verified via microsoft_docs_search 2026-02)"
},
{
"file": "skills/ms-ai-engineering/references/rag-architecture/semantic-ranker-reranking.md",
"text": "### Verified (MCP-research)"
},
{
"file": "skills/ms-ai-engineering/references/mlops-genaiops/genaiops-llm-specific-practices.md",
"text": "**Verifiseringsdato:** 2026-02-04"
},
{
"file": "skills/ms-ai-engineering/references/mlops-genaiops/model-versioning-registry-management.md",
"text": "### Microsoft Learn (Verified via MCP research, April 2026)"
},
{
"file": "skills/ms-ai-governance/references/responsible-ai/ai-risk-taxonomy-classification.md",
"text": "- **Metode:** MCP microsoft-learn server"
}
]
}
]
}