R1+R2 as one chunk — both KRITISK rows of the Q3 severity table sit on the
restore path, and neither closes alone.
R1: rollback-engine.mjs verified every checksum before AND after each write,
resolved the legacy backup root and reported createdNotRemoved — and none of it
was reachable. Measured: 16 files under scanners/ carry a process.argv entry;
the engine was not one of them. commands/rollback.md drove the restore as model
prose: an ESM import block a template cannot execute, ad-hoc `cp` offered
underneath as the runnable path, and "(checksum verified)" pre-rendered three
times in the success output. `cp` establishes no checksum, so the verification
was a property of the template rather than of the run — on the one surface that
runs when the user is already in trouble.
R2: implement.md Step 3 hand-built its backup (mkdir, cp, a date-derived id, a
manifest typed out in the template) while parseManifest knew one frozen sample
of that format, pinned by a HAND-WRITTEN fixture instead of by the template's
own text. Rename a key and parseManifest returns zero files while rollback
reports success.
Fixing only R1 leaves the new CLI parsing a prose format; fixing only R2 leaves
a clean format with no runnable entry.
- scanners/rollback-cli.mjs — --list / --create / --restore / --delete over the
existing engine, on the shared requireValidArgs gate. Exit 0 done, 1
outstanding (gate refusal with nothing written, or a backup that covered fewer
targets than given), 2 a file failed, 3 could not do the job. A gated restore
is 1, not 3: "this write leaves your project" is a verdict about a write that
WAS examined, and it rides in the payload where a command under 2>/dev/null
can act on it.
- createBackup gains `created` (recorded, never copied — no backup can hold a
file that does not exist) and `skipped`, so a backup covering fewer files than
asked is no longer indistinguishable from a clean one.
- implement.md Step 3 and rollback.md now call the CLI. parseManifest's
implement-format branch stays: nothing writes that shape now, but every backup
made before this chunk is on disk in it.
- backup-restore-contract.test.mjs checks every field rollback.md renders
against a payload produced by RUNNING the CLI. That is what replaced
"(checksum verified)".
20 guards seen red against the original state before any production code, then
each against its own defect. Two holes that surfaced there were mine: the
implement assertion matched `--create` as a substring of `--created` and stayed
green when the call was removed; and mutating the argv gate showed
requireValidArgs sets exit 3 by itself, so a CLI can report that it could not
parse its arguments and still run the restore underneath — that case is now
asserted on the bytes.
Suite 1752 -> 1777, 0 fail. Frozen tests/snapshots/v5.0.0 untouched. Dogfooded
through the templates' own command lines against a sandboxed HOME, including the
machine-wide arm: refused with the file unchanged, then restored under
--approve-scope.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Logq8GGWKhtyDem63FTEnG
New GAP finding CA-GAP-028: authored subagents exist and not one of them names
`model:` or `effort:`, so every delegated task runs on the main conversation's
model (`model` defaults to `inherit`). Cites BP-MODEL-001/002, landed in C1.
`whats-active` and `manifest` now carry `model`/`effort` per agent.
Shipped as a conditional LEVER rather than a 25th dimension, and the choice was
made by measurement: as a t3 dimension the agent-less marketplace-medium fixture
would count it vacuously-present, moving the denominators 41->42 and utilization
44->45 — which flips `segment` "Developing"->"Competent" in the frozen v5.0.0
posture baseline, a field strip-retired-gap.mjs does not mask. A lever never
enters those denominators. The general rule is now an invariant in CLAUDE.md.
One check across both axes, not one per axis: it fires only when neither is used
anywhere, so a deliberate everything-on-one-model policy stays silent. Cost is
recall, chosen for precision.
Found by dogfooding, fixed red-first: `model: inherit` is the documented default
spelled out, so it must not count as routing — otherwise a config opts out of the
opportunity without changing anything real.
Two pre-existing defects surfaced and closed on the way:
- The humanizer guard asserted TRANSLATIONS.GAP.static EQUALS the dimension
titles, which forbade humanizing any lever — all three existing levers fell
through to the generic "feature opportunity" default, wrong for a budget lever.
Guard now requires coverage of every emittable title, seen red against those
three before the entries were written.
- Two hand-written copies of the lever list (finding-codes guard, humanizer
guard) merged into one exported LEVERS registry carrying code AND title.
- suppression-validation pinned CA-GAP-028 as an unoccupied number; C4 claimed
it. Fixed structurally with a derived first-free id, not by picking a new
literal — same class as #60's "bump this again".
Suite 1596/0. Frozen v5.0.0 snapshots untouched.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pq3nye21RVYk4pZLeT8pGz
CC 2.1.226's /doctor Check 8 covers auto mode with usage-weighted judgement.
The binding positioning forbids carrying a feature whose whole value is
duplicating a /doctor check, so the "adopt this feature" nudge goes. The
deterministic side stays: SET still validates autoMode structure and still
flags it as dead config in shared project settings. GAP dimensions 25 -> 24.
The title lived in FOUR tables, not the two the removal was scoped against:
the dimension list, scoring TITLE_TO_ID, the humanizer's static translations,
and the scoring denominators (TIER_COUNTS t3 8->7, TOTAL_DIMENSIONS 25->24,
MAX_WEIGHTED 42->41) -- the one that moves a user-visible number. findGapId
falls back to 'unknown' silently, so a partial removal would have degraded
without failing. A blanket sync invariant now asserts all four against
GAP_CHECKS instead of comparing occurrences pairwise; each arm was verified
red against its own defect (denominator drift, orphaned humanizer entry,
resurrected dimension).
Frozen tests/snapshots/v5.0.0/ stays untouched. strip-retired-gap.mjs is the
removal twin of strip-added-scanner.mjs: it strips the retired dimension from
whichever side still carries it and re-derives GAP IDs, since retiring a
dimension from mid-list shifts every later ID by one. Derived utilization
figures are dropped from comparison rather than recomputed -- recomputing them
in a test helper would assert the new arithmetic against itself, and
scoring.test.mjs already pins them exactly. Re-seeding was rejected: it would
silently bake in any other drift across every scanner those four files cover.
risk_score, risk_band, verdict, overallGrade, maturity and segment are
byte-identical across the change (severity info carries zero risk weight; GAP
is excluded from the overall grade). Utilization shifts 43 -> 44 on the fixture.
D2 (CA-SKL-002) is NOT removed. Verified against the primary source first: the
CC changelog carries exactly one budget-fraction statement (L3786, 2.1.32) and
nothing supersedes it, so our 2% is current and 002 is not a duplicate with a
stale figure. /doctor's ~1% could not be reconciled from the changelog and it
discloses its own numbers as disk estimates, so it is recorded, not adopted.
Left explicitly unverified in a code note: L3786 says "character budget" while
we express tokens -- a 4x difference nobody can settle from the wording.
Suite 1531 -> 1535, all green.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RsfPGxgwbR3MY54wDC6hat
`KNOWN_OPEN` in cli-unknown-flag-rejection.test.mjs named two CLIs as still
carrying the argument-swallow defect. That number was the previous session's
field of view, not a measurement. Measuring all fourteen found **7** open on
the unknown-flag arm and **10** on a second arm the deferral note never
described.
Arm 1 — unknown flag: with no `else` branch, `--zzz` leaves no trace. exit 0,
full payload, a confident answer to a question the caller did not ask.
Arm 2 — the sharper one: `a === '--output-file' && args[i + 1]` asks only
whether a next token EXISTS, never whether it is a value. `manifest`,
`campaign-cli` and `knowledge-refresh-cli` each wrote a file literally named
`--json` into the caller's working directory when handed `--output-file
--json`, exit 0, with `--json` mode silently dropped. A wrong answer is bad;
an unintended file on disk is worse.
Two of the CLIs this catches were already in GUARDED and green on arm 1 while
arm 2 stood open a few lines away — the guard asserted one relation instead of
the invariant.
Fixed with a shared gate (`lib/cli-args.mjs`) that runs BEFORE each CLI's own
parse loop rather than replacing it: valid argv reaches the existing parser
byte-for-byte unchanged, so the byte-stability argument is structural rather
than empirical. `drift-cli`, `fix-cli` and `plugin-health-scanner` were
already correct on both arms and were moved into GUARDED instead of rewritten.
The three CLIs with a bespoke unknown-flag branch had it removed once the gate
made it unreachable.
Suite 1488 → 1531. Frozen v5.0.0 snapshots untouched; `self-audit
--check-readme` passed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014P6Rh59Mtj4uYrdYMCYZJE
Dogfooding `/config-audit plugin-health` against a fasit registered before the
run: 11 of 12 predictions confirmed, 1 refuted with evidence, 0 deviations.
The command's default path could not produce the report it documents.
M-BUG-21 (third arm): the argument loop ended in
`else if (!args[i].startsWith('-')) targetPath = args[i]` with no unknown-flag
branch, so `--output-file /tmp/x.json` was dropped and its value became the scan
target. Worse than in drift-cli: a non-existent path discovers no plugins, so the
scanner answered "No plugins found" (info) with exit 0 — a reassuring answer, not
an error. Unknown options and a value-less `--output-file` now exit 3.
M-BUG-33: the scanner had no `--output-file` and its default-mode report goes to
stderr, which `commands/plugin-health.md` discards with `2>/dev/null` before
telling the agent to read stdout. Zero bytes captured.
M-BUG-34: per-plugin rows and the grade formula never left `scan()` — the only
grade code, `formatPluginHealthReport`, had no caller — and cross-plugin findings
were flattened behind a `category` they share with per-plugin findings. The
mandated table and Cross-Plugin section were unbuildable, so the command had to
fabricate them. `scanDetailed()` now returns them; `scan()`'s frozen v5.0.0
envelope is unchanged by construction.
M-BUG-35: `.claude-plugin/marketplace.json` was flagged as an unknown file. It is
the documented catalog location, and `"source": "./"` makes the repo root its own
plugin, so one `.claude-plugin/` legitimately holds both.
Also: `commands/posture.md` ran both optional scanners in default mode under
`2>/dev/null` and read stdout — the same class as feature-gap.md:133 in the fix
chunk. A CLI-side flag fix does not close its callers.
Tests 1420 -> 1432, red first. Frozen v5.0.0 snapshots untouched.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XhhZ8zg1amR7YCAPqiZWdt
Every command so far asked an addition question — what to add, what to move,
what it costs. Nothing asked what is no longer earning its always-loaded rent.
This adds that axis as a fourth lensCheck on the existing hybrid motor rather
than a new scanner or a 22nd command: the measured payoff (~18% of one file)
justifies a mode, not machinery.
It is the only lens that proposes REMOVING config, so it carries a guarantee
the others don't need: a load-bearing block is never a candidate. Precision is
asymmetric — a missed dead line costs a few tokens per turn, a deleted one
costs a wrong remote or a broken script — so the floor is decided in code
(lib/floor-exclusion.mjs) before the opus judge sees anything, never in prose.
Granularity is the leaf block, with two structural exceptions: a paragraph
ending in ':' merges with the list it introduces, and an ordered list is a
contract whose steps inherit floor from any sibling. Unordered lists
deliberately do not inherit — a load-bearing bullet and a disposable one
routinely share a list, and container-reasoning is the error the hand-built
ground truth exists to catch.
Verified against that ground truth (built before any classifier existed), with
the comparison machine-checked rather than read by eye: zero load-bearing
blocks proposed, 11/18 deletable groups surfaced, ~756 tok ~ 18% of a ~4300
token file — inside the pre-registered band. The first run found five floor
violations the synthesized fixture missed; each got a structural rule and a
fixture shape so it cannot regress.
Three real bugs the dogfood run exposed, all now covered:
- JS \b is ASCII-only, so /\bunngå\b/ never matches — every Norwegian keyword
ending in æ/ø/å was silently dead.
- A bare word/word is not a path; "pros/cons" vetoed the largest deletable
block until PATH_RE was tightened to rooted paths and globs.
- "Mid-sentence" must key on a preceding lowercase letter; the loose version
read **bold labels:** and quoted openers as entities, costing 4 of 11 groups.
BP-SUB-001 is grounded entirely in the Anthropic steering blog already cited by
BP-MECH-001..004 and asserts nothing from the talk that motivated the feature —
no "80%", no ablation figure.
Suite 1365 -> 1382/0. Frozen v5.0.0 snapshots untouched; plain optimize output
byte-identical on identical input (--subtract adds keys only when passed).
knowledge-refresh-cli's reference date moved to 2026-08-01: its premise that
every seed entry was verified 2026-06-20 expired when BP-SUB-001 got a genuine
verification date, and backdating the entry to fit the test would have been a
lie about when its source was checked.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RW2haJXbxZpKivKHseSXNh
The GAP scanner's 25 presence checks ran over the full includeGlobal discovery, so
this plugin's own examples/optimal-setup (vendored across plugin-cache versions)
satisfied every tier-3 check — masking real feature gaps to GAP=0 on ANY target.
And the real ~/.claude/settings.json is invisible to the settings-key checks
(includeGlobal gotcha + maxFiles cap), which would flip statusLine/autoMode to
false positives once the maskers were removed.
- isAuthoredConfig: exclude plugin-bundled (~/.claude/plugins/) + nested examples/
and tests/fixtures/ (relPath-relative, so a fixture scanned AS the target keeps
its own files) from ctx.files + parsedSettings.
- readSettingsCascade: read the user->project->local settings cascade directly and
merge into parsedSettings — immune to the discovery cap/gotcha.
Empty target: ~0 (masked) -> 18 humanized opportunities; no statusLine/autoMode
false positives. Frozen v5.0.0 snapshots + SC-5/6/7 byte-stable (marketplace-medium
has no nested demo trees; hermetic-HOME cascade adds nothing). Suite 1350->1355/0.
Found by dogfooding feature-gap against the machine.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01683eAqVecv9VZfQzL8CQ9h
CPS originally inspected only files discovery classifies as claude-md, but a
CLAUDE.md can pull arbitrary files into the cached prefix via @import — and
those targets (e.g. @shared/conventions.md) are usually not claude-md in
discovery, so their inlined content was never scanned. Neither TOK Pattern A
(top-30 of cascade files) nor the in-file CPS scan reaches past the importing
file, so volatility inside an imported file was invisible.
B6 closes the gap: for each @import whose import site sits within the
cached-prefix window (imp.line <= CACHED_PREFIX_LINES), CPS resolves the path
(resolveImportPath, mirroring import-resolver/token-hotspots semantics), reads
the target, and runs findVolatileLines over its first 150 lines. A hit emits a
distinct medium finding — "Volatile content in @imported file breaks cached
prefix" — keyed on the resolved file, evidence naming the importer.
Scope boundaries (deliberate):
- One hop only; imports-of-imports stay with IMP (deep-chain owner).
- No lines-1-30 skip for imported content — that exclusion is root-file-specific
to avoid Pattern A overlap, which never reaches imported files.
- No double-reporting: an import resolving to a discovered claude-md is skipped
(own iteration); a reportedImports set dedupes a target imported by several
CLAUDE.md files.
Dropped from B6 (per plan verdict): confident behavioral cache-buster detection
(opusplan/model-switch is runtime, not static config) and jq-transcript
automation. "No overstated behavioral finding ships" — even the permitted
opusplan info-advisory was left out; the @import extension is the whole of B6.
Byte-stability: the in-file finding keeps the same condition + byte-identical
evidence/description (continue-skip refactored to if-emit, behaviour-preserving);
new findings fire only on a volatile import, which no frozen v5.0.0 fixture has.
docs: README + scanner-internals CPS rows + full B6 note; CLAUDE.md kept lean
([skip-docs]). Suite 1254 -> 1257 green; snapshots + SC-5 untouched.
Version/badges/CHANGELOG wait for the v5.10 release cut.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
HKV now flags hooks that build hookSpecificOutput.additionalContext from
un-grepped command output as an INFO advisory (weight 0, never severity-bearing
— excluded from the self-audit nonInfo set). That field enters Claude's context
every time the hook fires (plain stdout on exit 0 does not), so an unfiltered
payload is a recurring per-turn token cost.
- New lib scanners/lib/hook-additional-context.mjs: pure assessHookAdditionalContext
(unit-tested, no IO) + IO wrapper assessHookContextForRepo (walk hooks->scripts).
Heuristic: additionalContext + verbose-prone capture (cat/git log/execSync/…) &&
no filter (grep/head/jq/.slice). Deliberately low precision -> advisory only.
- HKV: emits the advisory inline on scripts it already reads (after the M5 verbose
check). Additive, info severity -> frozen v5.0.0 + SC-5 snapshots untouched.
- feature-gap: new filterHookLeverFinding companion, fires ONLY when >=1 chatty
hook is detected — surfaces the filter-before-Claude-reads lever (CC
filter-test-output.sh). Silent otherwise (opportunity, not noise).
- docs: README HKV + GAP scanner rows; scanner-internals.md HKV row + full B5
implementation note. CLAUDE.md kept lean ([skip-docs]); B5 fully documented in
README + scanner-internals.
Mechanism verified 2026-06-23 against code.claude.com/docs context-window.md
(additionalContext enters context; plain stdout does not). Suite 1239 -> 1254 green.
Version/badges/CHANGELOG wait for the v5.10 release cut.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
By default Claude Code defers MCP tool schemas (names-only, ~120 tok; full
schemas on demand via tool search). CA-TOK-006 detects config-file signals that
force the FULL schemas into the always-loaded prefix every turn:
- settings.json env.ENABLE_TOOL_SEARCH="false" (high)
- "ToolSearch" in permissions.deny (high)
- configured model is a Haiku model (medium)
- per-server .mcp.json alwaysLoad:true (CC v2.1.121+) (high)
auto[:N] is threshold mode (info, not a trigger).
New engine lib/mcp-deferral.mjs: pure assessMcpDeferral({settings,mcpServers})
(unit-tested, no IO) + thin IO wrapper assessMcpDeferralForRepo shared by TOK and
GAP. Severity scales with aggregate forced-upfront tokens (medium-confidence
reasons cap at medium). feature-gap cliOverMcpLeverFinding fires only as a
companion to CA-TOK-006 (prefer gh/aws/gcloud over MCP for common ops).
Honest scoping (Verifiseringsplikt): triggers on config files ONLY — never
process.env shell vars. Vertex / custom ANTHROPIC_BASE_URL / runtime /model
switch are launch state (would flap snapshots machine-dependently), so they are
DISCLOSED in every finding, not triggered. Tool-level anthropic/alwaysLoad and
claude.ai connectors likewise disclosed. Mechanism verified 2026-06-23 against
code.claude.com/docs (context-window.md, mcp.md#configure-tool-search +
#exempt-a-server-from-deferral, costs.md); the prefix-cache invalidation claim
was NOT-CONFIRMED in docs and is not asserted.
alwaysLoad added to CA-MCP VALID_SERVER_FIELDS (no longer flagged as unknown).
active-config-reader surfaces per-server alwaysLoad. Byte-stable: CA-TOK-006
fires only on new conditions; frozen v5.0.0 + SC-5 snapshots untouched.
Tests 1215 -> 1239 (engine 16, integration 5, lever 2, mcp-field guard 1).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
CLAUDE.md was 540 lines (CML >500 MEDIUM, config-grade B/89). Move the 19
per-scanner/per-block implementation notes (v5.6/v5.7 design rationale +
primary-source verification + byte-stability lessons) verbatim into
docs/scanner-internals.md under a new "Implementation notes" section, leaving a
read-on-demand pointer. Add a real "## Conventions" section pointing at
.claude/rules/ — this clears CA-CML-001 (missing recommended section), which the
moved detail headings had been masking via incidental rule/style keyword matches.
CLAUDE.md 540->134 lines / 8.5k chars; config-grade A/97; CML findings: none;
readmeCheck green; full suite 1168 green (hermetic).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
First detector of the 'is the config OPTIMAL?' axis (vs the existing 'correct?' scanners). New orchestrated scanner family CA-OPT (count 14->15), the deterministic half of the hybrid optimization lens.
CA-OPT-001 (low, Missed opportunity): a multi-step procedure in CLAUDE.md (>=6 consecutive numbered steps) that belongs in a skill. Reads recommendation + provenance from the best-practices register (BP-MECH-003). Conservative by design; the negative corpus proves null false-positives. Prose-judgment cases (lifecycle->hook, 'never'->permission) are deferred to the Chunk 2b opus analyzer.
Wiring mirrors OST: orchestrator entry, humanizer (OPT->'Missed opportunity' + family), scoring (OPT->'CLAUDE.md', existing area -> no new posture row -> byte-stable), strip-helper (OST,OPT), SC-5 regenerated under hermetic HOME (additive OPT entry only). 10 new tests; suite 1045->1055, self-audit A/A, readmeCheck passed (all verified with a clean HOME).
Note: the pre-existing TOK test reads the real ~/.claude (non-hermetic) -> run the suite with a clean HOME for deterministic results. Tracked as a separate follow-up.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
New orchestrated scanner output-style-scanner.mjs — first new family since
SKL. Three findings, each pinned to a CONFIRMED V-row of the steering-model
plan + re-verified against code.claude.com/docs/en/output-styles:
- CA-OST-001 (medium, V10): user/project custom style missing
keep-coding-instructions:true (default false) → silently strips built-in
software-engineering instructions when active. Scoped to user/project.
- CA-OST-002 (low, V11): plugin style with force-for-plugin:true overrides the
user's selected outputStyle. Verifiseringsplikt correction — the plan bullet
said "project/user style," but force-for-plugin is plugin-styles-only per the
docs, so the check keys on source==='plugin'.
- CA-OST-003 (medium): settings outputStyle matching no built-in
(Default/Explanatory/Learning/Proactive, case-insensitive) nor discovered
custom style → dead config.
Byte-stability — a scanner addition, not a field addition. Growing the
scanners array + scanners_ok cannot be hidden by a field strip, but re-seeding
frozen v5.0.0 (the SKL precedent) would now bake in B2's hotspot triple +
claudeMd drift. So, per the B2 lesson, frozen v5.0.0 snapshots are PRESERVED
and the OST entry is stripped at compare time via new
tests/helpers/strip-added-scanner.mjs (wired into json/raw-backcompat + the
Step 5/6 humanizer tests); only SC-5 default-output is regenerated (additive
OST entry, diff reviewed). OST is fixture-gated (no output styles on
marketplace-medium / hermetic HOME → silent).
Wiring: orchestrator; humanizer (OST→Configuration mistake) + humanizer-data
OST family (title-coupled); scoring (OST→Settings, keeps 10 areas). Suite
1012→1023 (+11). Badges: scanners 14, tests 1023, TRANSLATIONS families 15.
Lore swept: README, CLAUDE.md, scanner-internals, humanizer.md. self-audit A/A.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Annotate every ranked TOK hotspot with the load-pattern triple
(loadPattern/survivesCompaction/derivationConfidence):
- hotspotLoadPattern() maps each discovery `type` → a deriveLoadPattern kind.
Rules reuse activeConfig.rules for precise `scoped` handling; claude-md maps
by scope. Two new deriveLoadPattern kinds back the rest: `command`
(on-demand — body loads on /invoke) and `harness-config` (external —
settings/keybindings/.mcp.json/hooks.json/plugin.json configure the CLI, not
the model context, so they cost no per-turn context tokens). Honest split:
the .mcp.json FILE is external; the MCP server's tool schemas are a separate
`always` hotspot.
Byte-stability — the opposite of B1's manifest. token-hotspots IS a byte-equal
SC-6/SC-7 CLI, and its hotspots ride inside scan-orchestrator + posture, so the
change touched SIX frozen-v5.0.0 comparisons across five test files. Resolved by
preserving the frozen baselines: a shared tests/helpers/strip-hotspot-load-pattern.mjs
strips the additive triple before each byte-equal compare (proves the original
schema is byte-identical). SC-5 default-output snapshots (scan-orchestrator +
token-hotspots) regenerated — diff reviewed as additive-only.
Tests 1008→1012. Self-audit A/A, scanner count unchanged at 13 (C bumps to 14).
Completes v5.6 B (B1 manifest + B2 token-hotspots).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Consume the v5.6 Foundation enumeration in buildManifest:
- Component-level sources: drop the coarse `kind:'plugin'` roll-up (it
double-counted skills/rules/agents already enumerated once). Kinds are now
claude-md/skill/rule/agent/output-style/mcp-server/hook.
- Every source carries loadPattern/survivesCompaction/derivationConfidence.
Rules/agents/output-styles propagate the foundation-derived values; CLAUDE.md
maps scope→kind (all cascade files always-loaded); skills are tagged on-demand
(skill-body) so the body cost does not inflate the always-loaded subtotal.
- New `summary` (always/onDemand/external/unknown {tokens,count}); the
always-loaded subtotal — "tokens that enter context every turn" — is the headline.
manifest is an environment-aware CLI → SC-6/SC-7 verify it by mode-equivalence,
not byte-equal, and it is not in SC-5. Adding fields in place keeps all snapshots
green with no regen (verified). `total` changes (de-duped) — intended correctness fix.
TOK's load-pattern column (byte-equal SC-6) is deferred to the next chunk (B2).
Tests 996→1008 (deterministic buildManifest unit test + CLI presence checks).
Self-audit A/A, scanner count unchanged at 13.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Extends the DIS scanner and its shared permission-rules lib with two
documented Claude Code permission footguns. Verified verbatim against
code.claude.com/docs/en/permissions (fetched 2026-06-19).
- lib/permission-rules.mjs: new isIneffectiveAllowGlob(entry) — unanchored
tool-name globs in permissions.allow (`*`, `B*`, `mcp__*`) that CC silently
skips ("does not auto-approve anything"); valid only as a glob-free
`mcp__<server>__*`. Shared with CNF.
- lib/permission-rules.mjs: dominates() now treats the `Tool(*)` deny-all glob
as equivalent to a bare deny (covers a bare allow) — CC: "Bash(*) is
equivalent to Bash ... both forms remove the tool from Claude's context".
- DIS: new finding "Ineffective allow wildcard — Claude Code ignores this rule"
(low, permissions-hygiene, CA-DIS-NNN); the existing dead-allow finding now
also catches a bare allow killed by a Tool(*) deny.
- 9 new tests (5 lib, 4 DIS) + 2 fixtures (force-added past .gitignore .claude/).
Suite 903 -> 912. Snapshot unchanged, contamination grep clean. README/CLAUDE/
scanner-internals document the broadened DIS mandate; test badge synced.
self-audit: PASS, configGrade A 96, pluginGrade A 100, readme gate passed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ter3E2JSi1Khgmuf2kady8
Devil's advocate gap-verification (read-only Workflow, 9 skeptics) refuted the
blanket "all closed" claim by finding fix-cli.test.mjs was the one CLI-spawning
test still reading the real ~/.claude. fix-cli runs the SKL skill-listing
scanner (HOME-scoped) even with includeGlobal:false, so its manual findings
include CA-SKL-001 on a dev machine but not in clean CI.
This directly corrects 325182d, which listed fix-cli.test.mjs as "Proven safe,
left as-is (output byte-identical real vs empty HOME — fixable findings are
project-local HKV/RUL/SET, never SKL/COL)". That reasoning predated SKL being
wired into scan-orchestrator (7bb2547/66433fe) and was false: real HOME yields
manual=6 (incl. CA-SKL-001), hermetic manual=5 (5230 vs 4798 bytes).
- wrap all 5 fix-cli spawns in hermeticEnv() (matches the other 11 CLI tests)
- add a regression lock: a project-scoped run must surface no CA-SKL/CA-COL
- redirect the --apply backup check to HERMETIC_HOME — the test was also
writing backups into the real ~/.config-audit/backups on every run
- docs: stale "26 hook events" -> 28 (README:528, scanner-internals:73);
hook-validator.mjs comment April -> June 2026 (functional count already 28)
Re-audited all 12 CLI-spawning test files: 11 hermetic-helper, manifest custom
HOME-env, post-edit-verify safe-by-construction (early-exit only, never reaches
scanners). HOME-leak class now actually closed. Suite 875/875, no snapshot drift.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ter3E2JSi1Khgmuf2kady8
Syklus 2 of Fase 4 Items 2+3. Flags when the sum of active skill
descriptions exceeds the listing budget (~2% of context, CC 2.1.32).
Design (operator-confirmed "fact-first, 200k anchor"):
- low severity (estimate) vs medium for the verified 1,536-char cap
- each description counted up to the 1,536 cap (what actually loads in
the listing) — avoids double-counting the tail CA-SKL-001 flags
- fires when sum > 2% x 200k = 4000 tok; evidence leads with the measured
sum + a calibration note that the budget scales 5x on 1M-context models
- aggregate emitted after the per-skill loop so the common case reads
001=cap, 002=aggregate (finding IDs are a sequential counter, not stable
semantic IDs — tests match on title, never NNN)
Also:
- tailored humanizer static entry for the aggregate title
- fix latent HOME leak in posture-grade-stability.test.mjs: it spawned
posture.mjs without hermeticEnv(), so a real ~/.claude leaked HOME-scoped
SKL/COL findings into the baseline grade (Token Efficiency A->B). Now
isolated like the 8 other CLI-spawning tests.
- docs sync: test count 868->875, scanner-internals, gap-matrix, plan status
Suite 875/875, no snapshot drift, self-audit clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ter3E2JSi1Khgmuf2kady8
Fase 4 Items 2+3 (CC 2.1.114→181 gap-review). New orchestrated scanner
`skill-listing-scanner.mjs` (prefix SKL) flags every active skill whose
description exceeds the verified 1,536-char listing cap (CC 2.1.105, changelog
L1502). Past the cap, Claude Code silently truncates the description the model
reads to route skill invocation — dropping the trigger phrases at the tail.
HOME-scoped over all user + plugin skills via enumerateSkills (COL is the model).
- CA-SKL-001 (medium): description > 1,536 chars. Remediation folds in Item
2(b) — recommends disableBundledSkills + skillOverrides + trimming
(designvalg A: no standalone GAP-check, which would fire for nearly everyone).
- Designvalg B: v1 ships the verified cap ONLY. The aggregate 2%-of-context
listing budget is deferred — it needs a context-window assumption that would
turn a verified fact into a guess (would carry a CALIBRATION_NOTE if added).
- Choice C: recognize the skillOverrides settings key (CC 2.1.129) in
KNOWN_KEYS. Left OUT of TYPE_CHECKS — the value is a per-skill object
(off/user-invocable-only/name-only), not a string; a 'string' check (as the
plan sketched) would create a NEW false positive. Verify-first deviation.
Registration: scan-orchestrator (13th scanner), humanizer (SKL → 'Wasted
tokens' + static/_default translations), scoring SCANNER_AREA_MAP (→ Token
Efficiency; no 11th area), README badge 12→13, CLAUDE.md (finding-id +
test-count), docs/scanner-internals.md, gap-matrix + plan status notes.
Snapshots reseeded hermetically (SEED_SNAPSHOT/UPDATE_SNAPSHOT): SKL entry with
0 findings in empty HOME, scanners_ok 11→12, claudeMdEstimatedTokens bump from
the CLAUDE.md edits flowing through the cascade. Contamination grep clean.
Suite 868/868 (856 baseline + 11 SKL + 1 skillOverrides). RED→GREEN logged
per cycle.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ter3E2JSi1Khgmuf2kady8
Fase 4 token-opt, Item 1 of gap-review NEXT STEP #2. The prompt-cache pattern corpus + TOK scanner were frozen at an "Opus 4.7" framing after CC shipped Opus 4.8 (default, 2.1.154) and Fable 5 (2.1.170). Model-era facts re-verified against the official changelog cache before editing.
The patterns are properties of prompt-caching, not of any model, so mechanic text is now model-neutral with a single "current default: Opus 4.8" anchor — preventing a re-freeze at the next model bump.
- rename knowledge/opus-4.7-patterns.md -> prompt-cache-patterns.md (git mv, history preserved); 6 reference sites updated
- TOK scanner: line-318 finding text (human-facing) made model-neutral; header + cache-prefix-scanner + CLI comments refreshed
- configuration-best-practices.md body + footnote 4.7 -> 4.8
- human-facing docs: commands/{tokens,help,manifest}.md, project CLAUDE.md, README, docs/scanner-internals.md
- gap-matrix row marked DONE; future Items 2/3 retargeted to new filename
Failing-test-first (Iron Law): +2 knowledge staleness guards (era-anchor + no-refreeze) +1 scanner assertion (no stale model anchor in finding text). Suite 853 -> 856 green; zero snapshot drift; self-audit A(97) PASS. CHANGELOG / v5 plan / ratified gap-plan keep historical opus-4.7-patterns refs (correct record of past state).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ter3E2JSi1Khgmuf2kady8
`.mcp.json` has no per-server `trust` key — verified 2026-06-18 against
code.claude.com/docs/en/mcp + /settings. MCP server approval is
dialog/settings-based (enableAllProjectMcpServers / enabledMcpjsonServers /
disabledMcpjsonServers), never a JSON field. The scanner's "Missing trust
level" (CA-MCP-001, medium) and "Invalid trust level" (high) were false
positives flagging a field that does not exist.
- scanner: delete both trust checks + VALID_TRUST_LEVELS; drop `trust` from
VALID_SERVER_FIELDS so a stray `trust` is now flagged as an unknown field
- humanizer: remove the two trust-level entries
- knowledge (5 files): point to the real approval mechanism, not a trust field
- fixtures: scrub `trust` (incl. the invalid "local" in optimal-setup)
- tests: flip assertions (no trust-level finding; stray trust -> unknown
field) + add knowledge-staleness re-freeze guards
- snapshots: reseed (marketplace-medium .mcp.json -8 tokens, hermetic)
- gap-matrix: mark the trust verify-first item DONE
Suite: 853/853 green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ter3E2JSi1Khgmuf2kady8
Pre-trekexecute snapshot of in-progress CLAUDE.md/SKILL.md edits and
extracted docs/ files. Captured as one commit so /trekexecute claude-design
can run against a clean working tree.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
