Repository navigation
Fix(eval): Developer cases pr-1074, pr-1213 and pr-1320 state every behaviour their hidden tests assert - #1833
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub. |
|
VERDICT: APPROVE Judged head: 1b998ba Objectives version: 2f2439683458fec466dfb9c92a656a9d9a77b74f6fad9a7e150927711c3ea8e1 Ruling ordinal: 0 Brief hash: f6dfe0e4229053802fca801497ca4f7972bcc235f823eda64c8fdbef3047d16d Policy digest: bf51ddb9c7bd4776744d56ac91f4005ee8b2c1e4a5c50d1653eaa0ba8789accc Judged base: 9c30905 Manifest version: 2 Pause-bound ruling: (none) BRIEF CONFORMANCE: Brief conformance: the diff does what the brief asks. Three sentences are appended verbatim after one space to pr-1320, pr-1074 and pr-1213 in generic-task-statements.json. The case files carry the same text. Spec: apps/cli/specs/evals.md is unchanged, which fits a data-only re-cut with no behaviour change. Scope: only the seven surface-map files under apps/cli/tests/eval changed. There are no src, held-out or re-cut script changes. Tests: none were added and none were needed. I could not run the digest and re-cut checks because only the staged diff was available, so the digest values are unverified on my side. I checked that all four development files record the same new statements digest (888116f2…) and that the four DERIVED_FILE_DIGESTS development pins changed while the held-out pins did not. Only the developer and whole-loop files change any taskStatement text, and only for the three named cases. Docs: no doc owner bindings apply and the PR body is adequate. No findings. OBJECTIVES: FINDINGS (ordered by severity): SCOPE: (covered by the review summary) Tokens: 1831: review — Reviewer — claude — 6/1865/— |
|
VERDICT: PASS Judged head: 1b998ba Objectives version: 2f2439683458fec466dfb9c92a656a9d9a77b74f6fad9a7e150927711c3ea8e1 Ruling ordinal: 0 Brief hash: f6dfe0e4229053802fca801497ca4f7972bcc235f823eda64c8fdbef3047d16d Policy digest: bf51ddb9c7bd4776744d56ac91f4005ee8b2c1e4a5c50d1653eaa0ba8789accc Judged base: 9c30905 Manifest version: 2 Pause-bound ruling: (none) FINDINGS (ordered by severity): OBJECTIVES: CONFIG SCAN: (covered by the review summary) SECRETS: SECRETS: none found in my read of the diff. CI has not run (local review), so atta-labs/secret-scan is unconfirmed and I'm not claiming a scan pass. The only hex strings are SHA-256 digests of eval data files. Agent-config: not applicable. The PR changes no .claude/, .mcp.json or .agents/ files, so no scanner ran. The diff has no code, auth, dependency, injection or BYOK changes. It edits eval data JSON and pinned digests only: 3 task statements (pr-1074, pr-1213, pr-1320) in the statements file, mirrored into the developer and whole-loop case files, plus digest re-pins. I could not recompute the digests, because the repo checkout is not available here. No findings. Tokens: 1831: security — Security — claude — 4/1512/— |
|
No description provided. |
AEG provenance — task (backlog Issue #1831)
Tokens: 1831: archive — Archivist — — — — |
Closes #1831
For:
Claude Sonnet 5.5, coding-agent CLI on a dev machine, dispatched locallyProject: cli
Decisions
pr-1320,pr-1074andpr-1213entries ofgeneric-task-statements.json, and regenerated with the re-cut script instead of hand-editing any case file.DERIVED_FILE_DIGESTSfrom the regenerated files; held-out pins untouched.Test plan
Premise
Premise:
Evidence
Head: 1b998ba
Summary:
6 files changed, 17 insertions(+), 17 deletions(-)Group A — recomputable
git diff 9c309050e0281c5934bf24261803bca2fb6079ad...1b998ba1648c6f4e8171c1f98f855d6b884f9be8 --numstatGroup B — attested
vinaya check --all --diff-onlyGraded body: the live pull-request body (
--push)Group C — Test Plan commands
C1:
bun scripts/test-suites/local-run.ts exact --context 247c5e672700194088ff3264294d045e7ff04051a08745f0aad4dcc4119b7343 -- apps/cli/tests/checks/changeset-coverage-bin.test.ts@default apps/cli/tests/checks/changeset-coverage.test.ts@default apps/cli/tests/checks/check-exec-bits.test.ts@default apps/cli/tests/checks/own-workflow.test.ts@default apps/cli/tests/ci-shards.test.ts@default apps/cli/tests/eval/freeze-record.test.ts@default apps/cli/tests/eval/generic-suites.test.ts@default apps/cli/tests/eval/harbor-practice-fixes.test.ts@default apps/cli/tests/eval/harbor-tasks.test.ts@default apps/cli/tests/eval/step-one.test.ts@default apps/cli/tests/import-boundary.test.ts@default apps/cli/tests/lib/log-callers.test.ts@default apps/cli/tests/run-paths-only.test.ts@default apps/cli/tests/surface-spec-exports.test.ts@default harness/processes/vinaya-strict/tests/commands/task-status.test.ts@default scripts/package-compat/package-compat.test.ts@defaultReused from a green run recorded
2026-11-02T13:59:39.181Z(pre-push) — not re-run.C2:
bun apps/cli/tests/eval/recut-generic-cases.ts developmentReused from a green run recorded
2026-11-02T14:00:01.659Z(pr-report) — not re-run.C3:
git diff -U0 origin/main -- 'apps/cli/tests/eval/*-generic-cases.json' | grep -E '^[-+] +"taskStatement"'Reused from a green run recorded
2026-11-02T14:00:02.493Z(pr-report) — not re-run.Scope
Only the evals data under
apps/cli/tests/evalchanges: the statements source, the four development case files and their digest pins. Held-out split, re-cut script and every other statement untouched.Tier: 1