feat(verdicts): key each verdict on its own candidate, not the bundle's one IR projection (S3.2)

seed_store_from_bundle keyed EVERY `type: verdict` file on bundle_candidate_features — the single
candidate the bundle's validator-input.json describes. A bundle carrying verdicts about several
candidates collapsed them onto one key, so a verdict about candidate B scored a perfect structural
match against candidate A's query and could be folded into A's hypothesis prompt. The ExpeL
substrate was single-candidate by construction.

A verdict file may now carry its own structural key in frontmatter (affected_codes / measure_type /
claimed_saving_nok); absent, keying falls back to the bundle candidate, so every pre-S3.2 seed keeps
working unchanged. promote_verdict writes the three fields, so a promoted verdict — frequently about
a different candidate than the target bundle's projection — does not impersonate that candidate.

Semantics decided HERE, not pulled: commons' seeding rule (method-spec §3 Steg 1 + bundle example)
has not arrived; we said we would build locally first. D7 mirroring stays open.

- ALL THREE fields or none. A partial declaration raises VerdictFrontmatterError rather than merging
  with the bundle candidate, which would mint a key belonging to NEITHER candidate. Validation,
  never repair (mirrors write_concept_file); the tolerant-skip rule belongs to the RAW inbox layer.
- claimed_saving_nok parses via json.loads — the SAME literal rule the IR projection went through —
  and is written back with str() of the raw value. _mint_id hashes that value, so 30000 and 30000.0
  are different keys; a normalising writer would split one candidate's signal across two ids.
- The structural key is signal-free, so it does not weaken the Step-8 no-leak property (Test C green).

Load-bearing MEASURED, five mutations all red: detach per-verdict keying · detach the fields
promote_verdict writes · make a partial/unparseable key tolerant · normalise the magnitude on write ·
remove the fallback (control — breaks the step1 suite at collection, proving the fallback bears load).

589 -> 597 tests. Full gate green (pytest, ruff, mypy).

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QkjvTTxrg9LTrmghebfiij
This commit is contained in:
Kjell Tore Guttormsen 2026-08-03 16:44:58 +02:00
commit 012adc0a3c
10 changed files with 496 additions and 17 deletions

View file

@ -105,6 +105,25 @@ Python ≥3.10. MAF (`agent-framework-core` 1.9.0). Pakkehåndtering: `uv`. To b
(`tests/test_step8_promotion_loadbearing.py`): gaten avviser ikke-godkjent dom (RØD uten gate); godkjent
dom er navigerbar (RØD når `link_in_index` detaches); promotert signal holdes ute av `bundle_context`
(RØD når en beskrivende index-label lekker det inn). Index-RMW er ikke-atomisk (enprosess-MVP).
- **En dom nøkles på SIN kandidat, ikke bundelens ene IR-projeksjon (S3.2):** `seed_store_from_bundle`
leser hver `type: verdict`-fils EGNE strukturelle felt fra frontmatter (`affected_codes` /
`measure_type` / `claimed_saving_nok`); mangler de, faller nøklingen tilbake til
`bundle_candidate_features` — så hver pre-S3.2-seed står uendret (fallbacken er BÆRENDE: fjernes
den, brekker step1-suiten ved collection). Uten dette kollapset en bundle med dommer om flere
kandidater dem på ÉN nøkkel, og en dom om kandidat B scoret perfekt strukturell match mot
kandidat As query. **ALLE TRE felt eller ingen:** en delvis erklæring raiser
`VerdictFrontmatterError` i stedet for å slås sammen med bundle-kandidaten — sammenslåingen ville
myntet en nøkkel som tilhører INGEN av kandidatene. Validering, ALDRI reparasjon (speiler
`write_concept_file`); den tolerante hopp-over-regelen hører til RAW-innboks-laget.
`claimed_saving_nok` parses med `json.loads` — SAMME literal-regel IR-projeksjonen gikk gjennom —
og skrives tilbake som `str()` av råverdien, fordi `_mint_id` hasher den (`30000``30000.0`; en
normaliserende skriver ville splittet én kandidats signal på to id-er). `promote_verdict` skriver
de tre feltene, så en promotert dom ikke utgir seg for målbundelens kandidat; nøkkelen er
signal-fri, så Steg-8s no-leak-egenskap står. **Semantikken er bestemt HER** — commons' seeding-regel
(`method-spec` §3 Steg 1) hadde ikke kommet; D7-speiling forblir ÅPEN. Load-bearing MÅLT
(`tests/test_step32_multicandidate_loadbearing.py` + `test_step8_promotion_loadbearing.py`), fem
mutasjoner alle røde: detach per-dom-nøklingen · detach feltene `promote_verdict` skriver · gjør en
delvis/uparsebar nøkkel tolerant · normaliser magnituden ved skriving · fjern fallbacken (kontroll).
- **Globalt token-tak håndheves FØR kall, aldri bare etterpå (S3.4, F10):** `PortfolioBudget` +
`PortfolioMeter` er ÉN ledger over hele porteføljepasset (og — seedet av `read_spend` — på tvers
av pass), mens per-run `Budget`/`TokenMeter` er uendret. Taket har tre tenner, med hver sin jobb: