Repository navigation
fix(receipt): record checker images (PROOF-POLICY §8.2); legs in a pinned image - #57
Conversation
…nned image PROOF-POLICY §8.2 (owner ruling 2026-11-16, standards #1625) counts a witness toward PWG-1 only when its current receipt records the digest-pinned image its checker ran in. The receipt tools are now pinned at standards 940a27a, whose reader enforces that. - The current Agda row records debian:13-slim@sha256:a29215f6..., the container proofs.yml pinned for the Agda job at the run's own commit 7dfeab2. The Rocq and Lean legs of that run ran on the bare ubuntu-24.04 runner, which has no digest, so their rows carry none and every claim now reads PWG-2ii. README, PROOF-STATUS, the legs README and the proof-debt ledger say so and why. - The legs job runs in ubuntu:24.04@sha256:73b99939..., installing its tools as root. Both jobs pass --image, read from their own container: line at the commit under test, and --axioms none (the gates refuse any axiom), which the generator at 940a27a requires. The first run recorded on main will carry an image for all three legs and restore PWG-1. Local: --write and --check pass (the shared git checks, the pinned reader at PWG-2ii for N-1..N-3 and N-7..N-9, the negative control); --run-row with --axioms and --image writes a 13-field row for each leg. Not run here: the provers in their images; CI runs them. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BYbQRtNu216R13jtZLmoqs
|
🟢 clean · commit Advisory: 11 row(s), worst 🔴 high (0 critical, 9 high) — reported, never grading the change.
Hypatia scan: 20 finding(s)
panic-attack assail: 16 finding(s)
pons asinorum: 9 finding(s)
finishingbot (release readiness): 174 finding(s)
glambot (presentation quality): 95 finding(s)
rhodibot (RSR compliance): 13 finding(s)
accessibilitybot: not_applicable — template default: enrol once the repository serves HTML
cipherbot: not_applicable — template default: enrol once the repository implements cryptography
echidnabot: not_applicable — template default: enrol once the repository carries proofs and an ECHIDNA engine is configured
panicbot: not_applicable — panic-attack is already graded by the verdict grid's own panic-attack row, from static-analysis-gate.yml evidence; enrolling panicbot would report one scanner twice and give two rows that can disagree
seambot: not_applicable — template default: enrol once a seam register exists at spec/seams/seam-register.json
sustainabot: not_applicable — reserved fleet slot; no adapter exists (gitbot-fleet bots/sustainabot/README.adoc)
Scope: static findings reported by each row's tool at the named commit. The table grades only findings in files this change touches, so the colour answers "does this change introduce anything?", not "what does the whole repository still carry?". Standing findings are counted in each accordion and in report.json; the baseline filter remains the place where acknowledged debt is recorded. The lane state is the worst GATING row; advisory rows are listed and counted but never set it. Purple, red, yellow and blue grade the worst GRADED finding present; orange is evidence missing, malformed or a scanner that did not run; grey is an explicit scope exclusion with its reason. No row is a security certification. report.json, badge.json and the raw findings are in the verdict-grid artifact on the linked run. |
In the pinned ubuntu image, actions/checkout marks the workspace as a safe
directory only under a temporary HOME, so the next git call refused the
clone ("dubious ownership"; run 44214405363), as it did in the Agda job's
container before. The legs job now marks it for the job, as the Agda job
does.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BYbQRtNu216R13jtZLmoqs
|



Summary
PROOF-POLICY §8.2 now counts a witness toward
PWG-1only when its current receipt records the digest-pinned image its checker ran in. This comes from the owner ruling of 2026-11-16, landed as metadatastician/standards#1625.#54 pinned the receipt tools at standards
f6ebc5e, which predates that rule. This PR moves the pin to940a27a, standardsmain, and does three things:PWG-1.Until that run is recorded, the committed receipt reads
PWG-2ii. The Rocq and Lean legs of run 43455458225 ran on the bareubuntu-24.04runner. That runner has no digest, so under the current rule those witnesses cannot count towardPWG-1.Evidence from this PR's CI
Run 44214592597, on
1c28ec9, ran all three legs in pinned images:debian:13-slim@sha256:a29215f6….ubuntu:24.04@sha256:73b99939….PWG-1for N-1..N-3 and N-7..N-9.--checkatPWG-2ii.Changes
940a27a, each by sha256.debian:13-slim@sha256:a29215f6…, the containerproofs.ymlpinned for the Agda job at the run's own commit7dfeab2.proofs.yml.ubuntu:24.04@sha256:73b99939…, the image lintel#31 uses. It installs its tools as root and marks the workspace safe for git, as the Agda job does.--imageto--run-row. The value is read from the job's owncontainer:line at the commit under test, so there is one source of truth.--axioms none, which the generator at940a27arequires. The gates refuse any axiom.PWG-2ii, and why.After merge
Record the first run on
main, using the rows its receipt job prints. All three legs then carry an image, and the reader derivesPWG-1again.RSR Quality Checklist
Required
1c28ec9, including the Agda, legs and proof-receipt jobs. Locally,--writeand--checkpass.As Applicable
CHANGELOGupdated.Testing
🤖 Generated with Claude Code
https://claude.ai/code/session_01BYbQRtNu216R13jtZLmoqs