Agent belt for #4679 - #4680

Closes #4679

Source Issue #4679: Sync review finding on scripts/runner_lib/acceptance_gate.py: Preserve source-command matching when normalizing interpreters When a

Source: #4679

Why

sync-core reviewer chatgpt-codex-connector reported a P2 finding on stranske/Portable-Alpha-Extension-Model#2506, head fbd419c12df1, round 20261113T050423Z: normalizing /opt/anaconda3/bin/python -m pytest ... or python3.11 -m pytest ... to python3 -m pytest ... prevents failed gates from unchecking their original acceptance checkboxes, despite reporting 0/1 passed.

Evidence: stranske/Portable-Alpha-Extension-Model#2506 (comment)

Scope

Fix interpreter-normalization command matching at its source in stranske/Workflows, in scripts/runner_lib/acceptance_gate.py.

Non-Goals

Do not edit the generated consumer PR stranske/Portable-Alpha-Extension-Model#2506 or its sync/workflows-* branch. Do not touch other synced files or findings. No scaffolding or TODO-only changes.

Tasks

• Fix scripts/runner_lib/acceptance_gate.py to preserve source-command matching when normalizing interpreters.
  • Update acceptance_gate.py so each normalized command also keeps its original authored command text in a separate field (verify: confirm completion in repo)
  • Update the gate output consumed by renderDeterministicGates() so criterion matching can use the authored command as well as the normalized one (verify: confirm completion in repo)
• Add or extend a regression test beside scripts/runner_lib/acceptance_gate.py covering failed gates for criteria using /opt/anaconda3/bin/python -m pytest ... or python3.11 -m pytest ....

Acceptance Criteria

• Source-fix gates pass: node --test .github/scripts/__tests__/*.test.js for JavaScript and python3 -m pytest -q tests/ for Python; both must pass.
• Reintroducing the reported behavior makes the regression test fail; reverting that deliberate break restores passing results.
• The next .github/workflows/maint-71-merge-sync-prs.yml run reports chore: sync workflow templates Portable-Alpha-Extension-Model#2506 without this review thread, and sync-core stops bumping this issue.

Implementation Notes

.github/scripts/keepalive_loop.js uses criterion.includes(command.command) in renderDeterministicGates(). Preserve the authored command separately or support matching normalized commands.

.github/workflows/maint-68-sync-consumer-repos.yml regenerates the consumer sync candidate from stranske/Workflows. sync-core re-reads this issue each round and comments only when the sync PR head changes.

Original Issue

<!-- sync-core/v1 handoff fp=e18445238734 -->
## Why (verified evidence)
**review-finding** surfaced by sync-core on sync PR stranske/Portable-Alpha-Extension-Model#2506 (head `fbd419c12df1`), round 20261113T050423Z. Reviewer: chatgpt-codex-connector.

- Synced file: `scripts/runner_lib/acceptance_gate.py` (generated into the consumer by `.github/workflows/maint-68-sync-consumer-repos.yml` from `stranske/Workflows`).
- The finding, quoted:

> **<sub><sub>![P2 Badge](https://img.shields.io/badge/P2-yellow?style=flat)</sub></sub>  Preserve source-command matching when normalizing interpreters**
>
> When a checked criterion names `/opt/anaconda3/bin/python -m pytest ...` or `python3.11 -m pytest ...`, replacing its command text breaks the matching in `.github/scripts/keepalive_loop.js`'s `renderDeterministicGates()`: that function uses `criterion.includes(command.command)` to find checkboxes whose gates failed. The normalized `python3 -m pytest ...` is absent from those original criteria, so a failed gate leaves its acceptance checkbox checked while reporting `0/1 passed`. Preserve the authored command separately or update the renderer to match normalized commands too.
>
> Useful? React with 👍 / 👎.
>
> <!-- codex-code-review-finding:v1 -->

Evidence:
- https://github.com/stranske/Portable-Alpha-Extension-Model/pull/2506#discussion_r4559852804

## Tasks
- [ ] Fix the finding at its source in `stranske/Workflows`: `scripts/runner_lib/acceptance_gate.py` (the consumer PR stranske/Portable-Alpha-Extension-Model#2506 is generated from it by `.github/workflows/maint-68-sync-consumer-repos.yml` and must not be edited by hand).
- [ ] Add or extend a test beside `scripts/runner_lib/acceptance_gate.py` that fails on the behaviour the reviewer describes and passes after the fix.
- [ ] Let `.github/workflows/maint-68-sync-consumer-repos.yml` regenerate the sync candidate and confirm the review thread on stranske/Portable-Alpha-Extension-Model#2506 is resolved.

## Acceptance Criteria
- Named gate: `node --test .github/scripts/__tests__/*.test.js` for a JavaScript source, `python3 -m pytest -q tests/` for a Python source, run on the source fix; both must pass.
- Deliberate-break → revert: re-introduce the reported behaviour → confirm the named test FAILS → revert.
- Live: the next `.github/workflows/maint-71-merge-sync-prs.yml` run reports stranske/Portable-Alpha-Extension-Model#2506 without this review thread, and sync-core stops bumping this issue (it comments only when the sync PR head changes).

## Non-Goals
- Do NOT edit the generated sync PR stranske/Portable-Alpha-Extension-Model#2506 or its `sync/workflows-*` branch; `.github/workflows/maint-68-sync-consumer-repos.yml` regenerates it from the source.
- Do NOT touch other synced files or other findings; one finding per issue.
- No scaffolding / TODO-only changes; every task is a concrete edit verified by the gate above.

_Surfaced by sync-core round 20261113T050423Z; the quoted finding is verbatim from the evidence links and was present on head `fbd419c12df1`. sync-core re-reads this issue every round and comments only when the sync PR head changes._
<!-- workflow-source:local_request -->

Deferred Tasks (Requires Human)

• Let .github/workflows/maint-68-sync-consumer-repos.yml regenerate the sync candidate and confirm the review thread on chore: sync workflow templates Portable-Alpha-Extension-Model#2506 is resolved. (This task appears only in the original issue. The rewritten Tasks section correctly drops it. It would require triggering or observing a protected workflow and checking a thread in another repository. The agent cannot modify workflow files or retry CI pipelines, and it cannot verify the review state of an external PR. | Keep it out of the agent task list. Leave regeneration and thread resolution to the maintainers or to the automated sync run, and keep the live-verification criterion as a post-merge human or automation check.)
• The next .github/workflows/maint-71-merge-sync-prs.yml run reports chore: sync workflow templates Portable-Alpha-Extension-Model#2506 without this review thread, and sync-core stops bumping this issue. (This depends on an external workflow run and on behavior in another repository. The agent cannot trigger, retry, or observe it, and it cannot edit the protected workflow files. | Mark it as a post-merge manual verification step, not an agent-completable acceptance criterion.)
• Reintroducing the reported behavior makes the regression test fail; reverting that deliberate break restores passing results. (This is a process step that mutates the code, runs the tests, and reverts. It is feasible locally, but it is not a durable repo state that can be checked after the fact. | Convert it to a checkable property: the regression test asserts that a failed gate for a criterion containing /opt/anaconda3/bin/python -m pytest or python3.11 -m pytest is matched. The agent can report the break-and-revert result in the PR description.)

—
PR created automatically by the belt worker.

Source: Issue #4679

Closes #4679

Automated Status Summary

Scope

sync-core reviewer chatgpt-codex-connector reported a P2 finding on stranske/Portable-Alpha-Extension-Model#2506, head fbd419c12df1, round 20261113T050423Z: normalizing /opt/anaconda3/bin/python -m pytest ... or python3.11 -m pytest ... to python3 -m pytest ... prevents failed gates from unchecking their original acceptance checkboxes, despite reporting 0/1 passed.

Evidence: stranske/Portable-Alpha-Extension-Model#2506 (comment)

Context for Agent

Related Issues/PRs

• stranske/Portable-Alpha-Extension-Model#2506

Tasks

• Fix scripts/runner_lib/acceptance_gate.py to preserve source-command matching when normalizing interpreters.
  • Update acceptance_gate.py so each normalized command also keeps its original authored command text in a separate field (verify: confirm completion in repo)
  • Update the gate output consumed by renderDeterministicGates() so criterion matching can use the authored command as well as the normalized one (verify: confirm completion in repo)
• Add or extend a regression test beside scripts/runner_lib/acceptance_gate.py covering failed gates for criteria using /opt/anaconda3/bin/python -m pytest ... or python3.11 -m pytest ....

Acceptance criteria

• Source-fix gates pass: node --test .github/scripts/__tests__/*.test.js for JavaScript and python3 -m pytest -q tests/ for Python; both must pass.
• Reintroducing the reported behavior makes the regression test fail; reverting that deliberate break restores passing results.
• The next .github/workflows/maint-71-merge-sync-prs.yml run reports chore: sync workflow templates Portable-Alpha-Extension-Model#2506 without this review thread, and sync-core stops bumping this issue.