2 Commits

Author	SHA1	Message	Date
Edward Cheng-I Wu	75070eec84	feat(evals): #653/#828 add reviewer-calibration harness with isolated dispatch and audited scoring (#835) ... * feat(evals): #653 reviewer-calibration suite scaffolding — corpus assembler, isolated dispatcher, deterministic scorer, pre-registered rubric/RUN_PLAN (corpus freeze pending PDF access) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H2iNYa6YYYaPUwD2Z2Jr5e * feat(evals): #653 freeze the ICLR 2026 calibration corpus manifest (12 papers) + shared PDF-text normalization Corpus freeze (PR-A of #653): `corpus/papers.json` (label-free, 6+6 ICLR 2026 papers by the pre-registered seed; pypdf 6.11.0; pool hashes unchanged from the 2026-08-07 selection) and `manifests/gold_labels.json` (public Decision note ids + strings). No page-cap exclusion fired; `verify` PASS. First real-PDF contact found a hashing defect: pypdf emits lone UTF-16 surrogates from math fonts (61 in one sampled manuscript) and strict UTF-8 encoding raised, so `extracted_text_sha256` was uncomputable. The normalization now lives in one shared module (`scripts/_calibration_pdf_text.py`: NFC + lone-surrogate -> U+FFFD), imported by both the assembler and the dispatcher so freeze/verify/dispatch hash identical bytes; the rule is recorded in the manifest's `extraction.text_normalization` and `verify` fails hard on rule drift (a rule, not a version). Two tests added (41 total). `scripts/fetch_calibration_corpus.py` is the authenticated OpenReview operator tool that produces the freeze input, so the "third-party reconstruction" claim in the README is backed by a runnable path. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EehvYnmG8Xym5tXhTLfVm1 * refactor(evals): #653 simplify pass — shared hashing/fence/git-state, contract 1.1 docs /simplify findings applied (reuse, simplification, efficiency, altitude): - `_calibration_pdf_text.py` owns `sha256_hex` + `pdf_facts` (bytes hashed and parsed from one read via BytesIO; `extract_text=False` lets `verify` skip extraction when the pypdf version cannot be compared); surrogate replacement is one `re.sub` pass. Both the assembler and the dispatcher import it. - dispatcher reuses E4's closed data-fence grammar (`_delimited`), `_git_state` (declares unknown provenance dirty instead of raising), and the evidence path guard (`assert_plain_file`: rejects symlinked parent components, not just the leaf); one `_prepare` preamble for both stages; a text-hash mismatch now names its cause (installed vs manifest pypdf version). - assembler: exclusion rows stay dicts, `pool_list_mismatches` shared by freeze/verify, exclusion set built once. - scorer: `confusion`/`bootstrap_ci` take (predicted, gold) pairs (same RNG stream as before), `Counter` for the exact-mode vote, dead `_path` dropped. - RUN_PLAN/README: measurement contract 1.0 is closed to new rows (#664); the run publishes under 1.1 with its pre-registration record + write-once execution manifest (dispatcher/scorer support lands with the scored run). Re-freeze after the refactor reproduces papers[] and gold_labels byte-for-byte. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EehvYnmG8Xym5tXhTLfVm1 * fix(evals): #653 Iron Rule #7 at the two whole-file call boundaries + paper-id shape check Security review round 1 (first-party) found two below-threshold gaps and both are verified real: - The calibration dispatcher omitted E4's `DATA_BOUNDARY` sentence on the field-analyst call (the one E4 call that carries it, because `field_analyst_agent.md` states no untrusted-material rule of its own). Restored, and a fitted `REPORT_BOUNDARY` added on the synthesizer call, whose agent file is likewise dispatched whole with no such rule. Pinned by a transport-capture test that checks both sentences precede their fence. - Paper ids are spliced into file names (`<id>.pdf`, `cards/<id>/`) but `load_pool` accepted any non-empty string. Ids now must match `^[A-Za-z0-9_-]+$` (OpenReview's forum-id shape) in the assembler and the fetch tool; test pins the refusal. 43 tests pass. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EehvYnmG8Xym5tXhTLfVm1 * fix(evals): #653 codex round 1 — dispatch/verify invariant parity, card-path guard, scorer completeness Codex round 1 (gpt-6-astra xhigh) findings 2-7, 10, 11 and the cheap half of 9, each re-verified first-party before the change: - dispatcher: frozen cards go through the same plain-file guard as PDFs and agent files (a symlinked card1.md -> gold_labels.json was readable); the manifest's text_normalization rule and page_count are checked before dispatch, so dispatch and verify enforce the same manuscript invariants; transport-failure artifacts keep the partial stdout and stderr verbatim; every call attempt records RFC-3339 start/complete and prompt/output hashes into the panel record and cards freeze (the per-call evidence the heldout-measurement/1.1 execution manifest is built from). - verify: label must match decision_raw under the label transform; paper count and per-class label counts must equal the recorded quotas (synchronized paper+label removal no longer passes). - scorer: a second record for the same paper/replicate is a hard error, not a silent overwrite; a gold paper with no complete ensemble blocks the full tier; an A1 override needs its verbatim `raw` excerpt present in synthesis.md. Nine regression tests added (52 total). Real-corpus verify still PASS. Not addressed here (need a decision): finding 1 (camera-ready format leaks the accept label) and finding 8 (numeric seat scores vs categorical seat contract); finding 9's manifest/row builders land with the scored run. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EehvYnmG8Xym5tXhTLfVm1 * fix(evals): #653 drop the numeric score axis — protocol Phase 2 forbids AUC, seats are categorical Codex round 1 finding 8, verified against the source: the seat contract (eic/methodology/... agents) emits criterion-bound categorical judgements and states "Do not total, weight, average"; `calibration_mode_protocol.md` Phase 2 says "Do not report AUC: there is no continuous rubric score." The scorer nevertheless extracted a `Weighted Average` figure (a retired field) and RUN_PLAN promised AUC + score variance, so a conforming run would have published null numerics against a plan that promised them. The scorer now reports only what the protocol's full-tier table names: confusion matrix, balanced accuracy, FNR, FPR (bootstrap CIs), exact-label agreement (count/share/target-set size, with the binary-gold caveat), and replicate stability as categorical agreement (on side, on exact label). AUC is emitted as an explicit NOT REPORTED line. RUN_PLAN and the test fixtures follow. 52 tests pass. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EehvYnmG8Xym5tXhTLfVm1 * docs(evals): #653 mark the 2026-09-06 corpus SUPERSEDED (layout leaks the label, #828); RUN_PLAN model currency - README/RUN_PLAN: the frozen ICLR 2026 corpus is a harness-rehearsal corpus only — camera-ready replacement makes accepted PDFs visibly different from rejected submission PDFs (6/6 + 6/6; 30/30 in a fresh accepted-pool sample). No profile or measurement row may be published from it; the gold corpus becomes an ICLR 2027 submission-time capture. The "Why ICLR 2026" rationale is kept as pre-registered and annotated with the two facts that now cut against it (layout leak; Fable 5.1's 2026-06 cutoff covers the decisions). - RUN_PLAN + dispatcher default: subject `claude-fable-5` -> `claude-fable-5-1`, judge `gpt-5.6-sol` -> `gpt-6-astra` (provisional, #783 policy). Pre-dispatch edits, not amendments. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EehvYnmG8Xym5tXhTLfVm1 * feat(evals): #828 layout-tell guard at corpus freeze — refuse a corpus whose page-1 layout is not constant `assemble_calibration_corpus.py freeze` now reads page 1 of every cached PDF and evaluates four venue-template signals (published-as header, under-review header, "Anonymous authors", >=10 bare three-digit line numbers). Any signal that is not constant across the whole corpus refuses the freeze with the per-class counts; a uniform corpus records `layout_tell_check` in papers.json. `verify` recomputes the same check (skipped with a warning when a PDF is not cached; a manifest without the block warns). On the superseded 2026-09-06 ICLR 2026 corpus every signal is 6/0, so `verify` now FAILs on it by design. Shared `_open_reader` + `first_page_text` in the PDF helper. Six tests (signal detection, full and partial separation refused, uniform freeze + verify round-trip, missing-PDF skip, pre-check manifest warning). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014bJSoFkGeRaWTMJodk4VGh * feat(evals): #653/#828 rehearsal fixes + heldout-measurement/1.1 manifest and row builders Rehearsal 2026-09-06 (2 papers x 1 replicate, blocked at the first call by a rejected API key) exposed three dispatcher gaps, all fixed with tests: - credential preflight: zero-cost `GET /v1/models` before the first billed call; a definitive 401/403 refuses (key never echoed), network trouble is `inconclusive` and proceeds; outcome recorded in every record - credential rejection mid-run (`Failed to authenticate` / `API Error: 401` / `Not logged in`) is never retried (`CredentialRejected`); other transport failures keep the single retry - an aborted cards stage writes `runs/blocked-cards-<paper>.json` with its per-call rows instead of losing them; both stages share one record writer 1.1 contract substrate (RUN_PLAN "pre-registration record + execution manifest" item): - `dispatch_calibration_panel.py --stage manifest` folds the completed call rows of one attempt (frozen cards + panel records; `load_attempt` refuses mixed attempt identities) into a write-once, schema-validated `execution-manifest.json` - `build_calibration_measurement_row.py` composes the 1.1 row: plan and rubric hashed and compared against `frozen_commit` (drift refuses; dirty commit refuses), manifest re-derived from the records and compared field-for-field, judge rows required (no judges, no row), agreement recomputed by the checker's own `judge_divergence` (extracted from `check_heldout_measurement_report.py`, behaviour unchanged), validated by the checker before a write-once write - adjudication rubric gains `## Resolution direction` (flags_only, I13 lower-bound labelling); README tooling section; RUN_PLAN names the row builder; DATA_FLOWS names the dispatcher's preflight touchpoint; scorer docstring de-staled (no score axis); pytest manifest +1 No calibration number is recorded anywhere in the repository. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014bJSoFkGeRaWTMJodk4VGh * fix(evals): #653/#828 codex round 2 — bind every row input to its attempt, harden the guards 12 of 13 findings applied (gpt-6-astra xhigh, read-only exec): - P1 foreign metrics: scorer output is bound to the attempt (per_panel keys == the complete panel records here, attempt ids match, n_papers matches) - P1 raw drift: record admission re-hashes every completed call's raw output against output_sha256 (manifest stage and row builder alike); prompts are not retained (they embed the manuscript) - P1 preflight redirects: the probe uses a no-redirect opener (a 3xx is `inconclusive`) and skips a non-https ANTHROPIC_BASE_URL - P2 estimand: class-A adjudication is now pre-registered as bidirectional (every synthesis decision transcribed blind and compared with the grammar), so the row publishes a point_estimate instead of an I13 "lower bound" that only meant audit coverage - P2 pre-write parity with R5: manifest timestamps parsed and ordered before the write; declared claims checked against the local manifest - P2 strict JSON: inputs parsed with the checker's strict loader, outputs serialized with allow_nan=False and round-tripped - P2 judge failures: `--blocked-run` ledger entries merge into attempts.blocked_runs (I11) - P2 admission by content: suite/stage/status/provenance from the record body, never the filename; blocked records are identity-checked too - P2 cards re-run: a reused evidence dir refuses (write-once stage records) - P2 auth signature: anchored at the start of stdout/stderr and limited to exit-code failures; a timeout's partial prose is never a credential error - P2 layout signals: phrase tests run on whitespace-folded text - P2 partial PDF cache: verify checks every cached PDF (can refuse, cannot clear) instead of skipping the guard - P3 real `git show` test for sha256_at_commit on a temporary repository Partially applied: "distinguish unobservable signals from absence" (not built; the constancy rule is pre-registered as stricter by design). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014bJSoFkGeRaWTMJodk4VGh * fix(evals): #653/#828 shared transport — capture every assistant message, fence the subject's config Rehearsal take 2 (2026-09-06/07, 8 billed calls on the first paper) found two transport defects in `ClaudeCliTransport`, shared by the E4 and the calibration dispatchers: - text-mode `claude -p` prints only the LAST assistant message: the first paper's synthesis (long enough to be continued) came back starting mid-table, with the Editorial Decision Letter and its `### Decision:` line in the missing head. The transport now runs `--output-format stream-json --verbose` and concatenates the text blocks of every assistant message; an error result or an unreadable stream is a TransportFailure that keeps the raw bytes. - `--bare` does not fence the subject: a two-call probe on 2.1.260 showed the operator's whole global CLAUDE.md arriving as a system-reminder, plus `settings.json` `language` and the output style (the seats appended Traditional-Chinese "plain-language summary" sections). The subject now runs with an allowlisted environment (PATH/HOME/LANG/TMPDIR/TERM/USER/ SHELL + ANTHROPIC_*; no CLAUDE_* inherited from a parent session) and a per-transport empty `CLAUDE_CONFIG_DIR`; the same probe then reported no instruction beyond the SDK identity line and the date. E4 tests: one fake updated to emit stream-json; five new tests (message joining, error/junk results, unreadable-stream failure with bytes, environment allowlist, argv/env of a live call). Calibration docs and the panel record's `dispatch` field describe the new recipe (pre-dispatch change, no amendment). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014bJSoFkGeRaWTMJodk4VGh * fix(evals): #653/#828 codex round 3 on the shared transport — eviction signals, LF framing, network env, failure evidence Five P2 findings (gpt-6-astra xhigh, read-only exec), all applied: - refusal-fallback eviction: assistant `supersedes` and system `model_refusal_fallback.retracted_message_uuids` (wire fields verified in the installed CLI 2.1.260) drop retracted partials before concatenation - NDJSON split on LF only (`str.splitlines` also splits on U+0085 / U+2028 / U+2029 inside a JSON string); CRLF tolerated - environment allowlist keeps documented network/TLS inputs (proxies, NODE_EXTRA_CA_CERTS, SSL_CERT_*, CLAUDE_CODE_CLIENT_*); an apiKeyHelper that needs more is documented as unsupported behind the fence - transport failures carry assistant TEXT in `stdout` and the raw stream in `raw_stdout`; a framing-only stream is "no model response" (E4 no longer writes stream metadata as a partial response); both dispatchers preserve the raw stream as `*.transport-stream.jsonl` - a structured error result (`[TRANSPORT: result <subtype>]`, diagnostic in stdout) is classified by the calibration retry loop like the plain-text startup failure: a credential rejection is never retried E4 tests +6 (256), calibration +1. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014bJSoFkGeRaWTMJodk4VGh * feat(evals): #653/#828 keep the raw stream of successful calls as evidence `ClaudeCliTransport.last_raw_stdout` exposes the stream-json framing of the most recent successful call; the calibration dispatcher writes it next to the text as `<label>.transport-stream.jsonl`, so the next rehearsal shows how many assistant messages a deliverable spanned (the 2026-09-06 synthesis lost its head to exactly that). Probe 2026-09-07: a 12,000-line reply at effort low arrived as ONE text message after a thinking-only message, so the head loss is attributed to multiple text messages in one turn (likely interleaved thinking at xhigh), not to an output-length continuation; the parser covers both. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014bJSoFkGeRaWTMJodk4VGh * fix(evals): #653/#828 allow requiring a successful credential preflight * fix(calibration): bind audited decisions and preserve failed dispatch evidence * fix(transport): retain truncated UTF-8 output as byte evidence --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>	2026-09-08 08:38:58 +09:00
Edward Cheng-I Wu	cdd48d916d	docs: DATA_FLOWS.md — single map of network touchpoints + local stores (#758) (#770) ... * docs: single data-flow map + DF-1..DF-3 coverage lint (#758) Add docs/DATA_FLOWS.md — one row per network touchpoint (trigger, payload class, recipient, credentials, off switch) and one row per local store (path, content, TTL, deletion), with an explicit scope statement (the Claude session itself is platform-governed; nothing publishes autonomously). Covers the four gate resolvers, the standalone Chinese-literature resolver (NOT in the gate), the consent-bound claim-standing discovery adapters, both cross-model transports (API and citation-only Codex subscription), the SessionStart update check, the manual smoke tests, and the v3.9.4 timeline bootstrap — the last one surfaced by the new lint itself on first run (it was absent from the #758 issue enumeration). Inbound links from README, SECURITY.md (in-scope exfiltration anchor), and THIRD_PARTY.md (core-suite vs third-party contrast). Lint (same-PR drift-point discipline): scripts/check_data_flows.py — DF-1 every non-test scripts/*.py importing a network module (AST scan, so no-call guards naming urllib.request in strings do not count) must be named on the map; DF-2 same for curl-invoking shell scripts; DF-3 README/SECURITY/THIRD_PARTY keep a rendered resolving inbound link (fences + HTML comments stripped with the semantics converged in the PR #768 review; consolidation into a shared helper is follow-up). 12 mutation tests; wired into spec-consistency.yml + pytest manifest. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EosnA4RdUYgbF2KmZ1DTmc * refactor: apply /simplify pass (4-agent, deduped) (#758) Doc: the four gate resolvers collapse into a 4-column sub-table under one shared trigger/payload/off-switch lead (the wide table kept only heterogeneous touchpoints); the exhaustiveness sentence is bounded to what DF-1/DF-2 actually detect (direct imports + curl; spawned-CLI and session-tooling paths held by review); "Nothing here publishes" now inherits POSITIONING.md and its not-a-runtime-guarantee qualifier; the subscription-free note is trimmed to its rationale; Related gains the SETUP bullet as the tunables authority. Coverage: docs/SETUP.md becomes the fourth DF-3-pinned inbound surface (pointer added in the cache section); the four translated READMEs mirror the README pointer; docs/DATA_FLOWS.md registers into check_spec_consistency.py relative-link validation. Lint: DF-1 module vocabulary rebuilt as the network subset of the no-call envelope FORBIDDEN_IMPORTS (deviations documented: dotted urllib.request/http.client instead of bare urllib/http; ssl excluded); scan is now recursive into scripts/ subpackages. Tunable constants in verification_cache.py gain update-both comments. Tests: the three hollow assert-baseline tests become real mutations (name-based test exemption, uncomment-curl, from-urllib idiom); recursive-scan and SETUP-surface tests added (17 total). Skipped with reason: endpoint-hostname lint (near-zero event rate, composed-URL false-fire risk); row-id shrink constant (review-owned per degradation-registry precedent); markdown-helper consolidation with check_control_availability.py (whichever PR merges second extracts the shared module — recorded in both PR bodies). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EosnA4RdUYgbF2KmZ1DTmc * fix: close codex R1 findings — 8 P2 + 4 P3 (#758) Doc accuracy (6): Chinese-literature row rewritten (callable client, no CLI; PubMed path sends the required NCBI contact email + bibliographic search coordinates); codex-transport payload names citation_context (can contain unpublished manuscript text); update check documented as one curl transfer per 24 h with redirects and the ARS_UPDATE_CHECK_REMOTE_URL override; retraction-status SQLite cache added to local stores (caller-supplied path, 30-day stale threshold, no auto-expiry); discovery adapters credentials corrected (fixed User-Agent, resolver env keys not consumed); resolver payload narrowed to identifiers + title query strings; update-check state content corrected (state label + two version strings). Lint mis-pass/mis-fire (4): DF-1/DF-2 coverage now requires the full repo-relative path (basename-substring collision closed); DF-2 is recursive over scripts/ and hooks/, recognizes path-qualified curl, and masks quoted spans before the comment strip; DF-3 strips inline code spans before link extraction (a backticked link does not render). Five mutation tests added (22 total). The code-span rule is a divergence from check_control_availability.py to be carried over at the declared helper consolidation. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EosnA4RdUYgbF2KmZ1DTmc * fix: close codex R2 findings — 4 P2 + 1 P3 (#758) - DF-2 scans command-substitution bodies BEFORE quote masking, so resp="$(curl ...)" — a real network call inside double quotes — fires (mutation test added; suite now genuinely 22, correcting the prior commit message which said 22 when 21 were collected). - Map gains the Codex audit wrapper row (scripts/run_codex_audit.sh: human/CI/hook-invoked only, sends deliverable + supporting file contents through the local Codex CLI login). - Update check re-bounded: at most one SUCCESSFUL check per 24 h; a failed attempt writes no state and may retry next session. - Cache TTL wording corrected: expiry is a cache miss, not deletion; expired rows persist until invalidated or the file is deleted. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EosnA4RdUYgbF2KmZ1DTmc * fix: full comment lines execute nothing — DF-2 substitution scan (#758) The R2 command-substitution scan ran before any comment handling, so a full comment line containing $(curl ...) false-fired — surfaced by the codex R3 pass (timed out mid-review, but its transcript had already demonstrated the false fire). Comment-only lines are now skipped before the substitution scan; a $(curl) inside a trailing inline comment remains a documented accepted edge. Mutation test added (23 total). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EosnA4RdUYgbF2KmZ1DTmc * fix: close codex R3 findings — command-position curl + image links + retention wording (#758) - DF-2 rebuilt around COMMAND POSITION: curl counts only as the first non-assignment token of a segment (pipes/separators/substitution openers), so `command -v curl` preflights and `echo curl` no longer false-fire; VAR=x curl still fires; wrapper-prefixed invocations (sudo/timeout) are documented accepted edges. - DF-3 link grammar excludes image syntax —  renders no anchor and cannot keep the acceptance criterion green. - Cache retention wording includes the overwrite path: expired rows persist until overwritten by re-verification, invalidated, or the file is deleted. - Three mutation tests added (26 total). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EosnA4RdUYgbF2KmZ1DTmc * fix: close codex R4 finding — curl behind shell control words (#758) The command-position head-token scan now skips shell control words (if/elif/while/until/then/else/do/!/time/exec) before naming the head, so `if curl …; then` and `while ! curl …; do` fire while `if true; then` stays quiet. Two mutation tests (28 total). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EosnA4RdUYgbF2KmZ1DTmc * fix: close codex R5 finding — option tokens after control words (#758) `time -p curl …` / option-bearing exec forms: the head scan now skips `-`-prefixed option tokens alongside assignments and control words, so the option cannot shadow the command head. Mutation test added (29 total). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EosnA4RdUYgbF2KmZ1DTmc * fix: close codex R6 finding — harness spawned-CLI paths scoped out (#758) Three maintainer-only measurement scripts reach the network through locally authenticated CLIs (dispatch_e4_panel via claude -p, run_review_criteria_constructive_value via Codex, check_ranking_lift via gh api). They are not user-facing feature paths, so instead of diluting the touchpoint tables they are now an explicit named scope exclusion — the exhaustiveness claim no longer silently spans them. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EosnA4RdUYgbF2KmZ1DTmc * fix: close codex R7 finding — boundary count wording (#758) "Two boundaries" became three after the R6 harness exclusion; the count is removed rather than maintained. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EosnA4RdUYgbF2KmZ1DTmc --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>	2026-08-18 00:28:24 +08:00