Five places still described the pipeline as it was before code that is
already in dev:
- the S3 hint of the task packet told the author to push the branch, while
the spec lives in the issue body (§2.3, #517) and nothing is pushed
before S5 (§11.8);
- process-gate printed «FAIL п.9 Gates: light» for a trailer nobody writes
or reads, while §10.2 item 9 is the unimplemented release:prerelease
verdict check. The check is removed; a contract test ties every RULES key
to an implemented item of §10.2 and every finding number to a RULES key;
- §10.4 item 4 demanded a heredoc in run:, while #723/#730 and their tests
demand the opposite: commit messages echo line by line into a file,
comment and summary texts come from code;
- the ship merge comment, AUTHOR.md, REVIEWER.md and AGENTS.md named only
the pre-beta document, though since #727 the night reads ship code first;
- the nightly publication committed «docs: ship review for nightly …
перед бетой» with the beta step's Issue: #696. It now has its own
subject (the document name), body and Issue: #727; the beta message is
unchanged.
The browser-guard inventory note still said growth above 200 fails
mutation-gate --check; since #699 it is a guideline and --check warns. Its
counts now match the inventory: 205, lifecycle 90.
Issue: #748
User-Visible: no
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
The body of _process.yml is read from dev (@dev, #623), so the flags and
formats it passes to scripts are dev's. After "Опубликовать документ ревью"
the working copy of job integrate is the task branch, and a show/ship branch
with a clean merge is not rebased before review: its scripts/ may lag dev by
days. review-doc-guard.mjs silently ignores unknown flags (the anchor lost
#726 route and #737 usage), and a stale merge-candidate.mjs merges the old
way. Only two calls (#723 push refusal, #726 route) were taken from dev, each
with its own extraction, and on ship/reuse the remaining ones ran dev's
version anyway: the script version depended on the path.
Now one step right after setup-node extracts
`git archive origin/dev scripts .github/workflows/validate.yml` into
$RUNNER_TEMP/dev-tools and every repo script of the job runs from there via
TOOLS (review-result-gate, review-doc-guard, reviews-index, merge-candidate,
process-track route, status-label). validate.yml is part of the snapshot
because workflow-jobs.mjs reads it relative to itself; without it ci-proof
answers `failed (#622)` and every code merge would return to S6. The working
copy stays the material: git, the document and paths are judged there.
PROCESS.md §10.4 gets the paragraph "Скрипты конвейера — из dev": the
model_review exception, merges of pipeline changes judged by dev's version,
and compatible edits of the Validate proof contract.
Tests: test/process-integrate-tools.test.mjs is the job contract (no step
calls scripts/ from the working copy, every call goes through the snapshot,
one archive from origin/dev with validate.yml, and the step as is yields a
directory where ci-proof resolves the job contract); publish-push-refusal
runs the publish step and the #413 step on real bash with a task branch whose
review-doc-guard.mjs exits 7 (red with the old call). Existing harnesses take
the snapshot step before the publish and decide steps; the #706 mutant anchor
follows the status-label call.
Issue: #749
User-Visible: no
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
The weekly process metrics weigh tracks and the nightly ship review in the
order quality, speed, tokens (#707), but the third axis had no source: the
claude-code-action step hides usage from the Actions log on purpose, nothing
read its execution_file, and the #728 reader printed "no data" every week.
scripts/model-usage.mjs is the single module that builds and parses the line:
`<!-- hp:usage input_tokens=N output_tokens=N cache_creation_input_tokens=N
cache_read_input_tokens=N num_turns=N -->` (sums over every model in the last
`result` message, `result.usage` when modelUsage is absent) or
`<!-- hp:usage-none reason=<code> -->`. Only the result message is read; the
rest of the file holds tool results, and no byte of it is printed.
A new step right after Review in both model_review jobs (always(),
continue-on-error) hands the line out as the job output `usage`. Usage is a
reporting figure like the stage duration, so it travels as a job output and
not through the sealed artifact: REQUIRED_FILES and the #556 gate are
unchanged. Publication treats the line as untrusted input and writes the
normalized form as the last line of the anchor block (review-doc-guard
--anchor --usage=) or right after the SHIP-REVIEW block; empty becomes
reason=missing, anything off-format reason=invalid.
The #728 reader now takes the line only from the machine block: a reviewer
quoting the previous round in prose no longer doubles its usage, and
"no data" is counted as missing, never as zero. PROCESS.md §10.4 documents
the source, the format and why it is a job output.
Issue: #737
User-Visible: no
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
Review r1 (Medium): refusalSummary said "повтор и ребейз не помогут" for every
non-stale outcome, and since AC2 the rebase guard's summary carries it too.
For a workflow-permission refusal a rebase and push by the author is exactly
the way out (PROCESS.md §10.4). That outcome now says so; other GitHub
refusals keep the old sentence.
Issue: #730
User-Visible: no
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
After #705 and #723 two more workflow bodies still treated every failed
push as a moved dev: the SHIP-REVIEW publication (_ship-review.yml) retried
three times with "dev went ahead", and the derived-artifacts bot commit
(_beta-derived.yml) told the release manager to rerun the workflow. A
refusal by GitHub itself - a token without the workflow right, a branch
rule, a hook - is cured by neither, and neither step said what GitHub
answered.
Both pushes now keep stderr and hand it to the #705 classifier through the
same CLI (merge-candidate.mjs --push-refusal). A stale lease keeps the old
behaviour: another attempt for the ship review, the rerun advice for the
derived artifacts. Any other outcome stops the step at once: the log gets
the git answer and the step summary gets the reason and the git answer
without secrets (--summary, refusalSummary with the new ship-review and
beta-derived labels). The classifier comes from dev, as for the other steps
of these bodies: both jobs check out dev, and the ship review resets to
origin/dev before every attempt. The ship review commit message is built
line by line into a file instead of a heredoc, as in #723. The thin callers
ship-review.yml and beta-derived.yml are untouched.
The rebase guard in _process.yml also writes the refusal reason to its step
summary now (--summary, label "rebase"); a stale lease writes none.
test/publish-push-refusal.test.mjs runs both steps as they are with real
bash and real git in temporary repositories (moved dev = a real neighbour
push, GitHub refusal = recorded stderr with a token, a credential URL and
an Authorization header); on the old bodies 10 of its 12 new tests fail.
The #705 execution tests of the rebase guard in rebase-generated.test.mjs
now also read the step summary. PROCESS.md names the two steps next to the
Issue: #730
User-Visible: no
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
A non-green show verdict that found "something to decide" went down the same
path as "fix the code": S6 with a limit of 2. Promoting the task to track:ask
was left to the agent's memory, with no named criterion and no trace, and the
exhausted budget only surfaced on the next S7 - after a fix nobody would read.
The structured verdict now carries `route` (fix | reclassify) and an optional
`criterion` (one of the six show criteria of PROCESS.md section 5). The trust
boundary reads a missing route as fix, rejects one outside the dictionary and
rejects reclassify on a green verdict. `reviewRoute` in process-track.mjs is
the single decision: on a code review of an unconfirmed show it moves the task
to track:ask and S3-spec; on an owner-confirmed show it adds `blocked` and asks
the owner; anywhere else reclassify degrades to fix with a note. The verdict
that spends the last cycle sets review-4 at once; the stage budget is shared
across tracks, so promotion changes the limit (4), not the count.
The "Решение по вердикту" step makes one `process-track.mjs route` call (from
dev, like the track step) and only executes its output: comment from a file,
labels from add/remove lists, status via status-label.mjs as before. The track
step also emits `confirmed` and a `route_note` for the review prompt; the
review document anchor gains a route tail that the old reader still parses;
wait-verdict reports the two new pipeline comments. The guard's own
spent >= limit check stays as the safety net.
Issue: #726
User-Visible: no
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
Two steps publish a commit and treated every failed push as a moved branch:
the release review job (release-review.yml) retried three times with "dev
went ahead", and the review document step (_process.yml) rebased and pushed
again. A refusal by GitHub itself - a token without the workflow right, a
branch rule, a hook - cannot be cured by a retry or a rebase, and the step
never said what GitHub answered.
Both pushes now keep stderr and hand it to the #705 classifier through the
same CLI the rebase guard uses (merge-candidate.mjs --push-refusal). Only a
stale lease (rejected / fetch first / stale info) keeps the old retry or
rebase. Any other outcome stops the step at once, without retries: the log
gets the git answer and the step summary gets the reason and the git answer,
both passed through redactSecrets (token, credential URL, Authorization).
The review document step takes the classifier from dev, as the rebase guard
does: a task branch behind dev may not carry it.
The summary text is written by the new --summary option (refusalSummary),
not by a multi-line string in run:, and both commit messages are now built
line by line into a file instead of a heredoc (PROCESS.md §10.4 item 4).
release-review.yml is dispatch-only and is not mirrored to main. PROCESS.md
names the rule next to the rebase guard; the #638 trailer witness in
test/release-review.test.mjs follows the line-by-line message.
test/publish-push-refusal.test.mjs runs both steps as they are with real
bash and real git in temporary repositories; only the push transport is
replaced: a moved branch is a real neighbour push, a GitHub refusal is a
recorded stderr carrying a token, a credential URL and an Authorization
header. On the old steps 9 of its 11 tests fail.
Issue: #723
User-Visible: no
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
