prompt: route smoke-runner-delegation (#260) and record the shape decision - #280
Merged
Merged
Conversation
Re-measured all 10 run_smoke.py copies from every repo's main before
planning, and recorded the decision the prompt gates on.
The 2026-08-05 correction is itself stale. Measured today: 10 copies, 3
variants, ZERO functional drift inside any variant — workspace (356L)
byte-identical across its 3 repos, workspace_test (198L) diverging only in
2 docstring lines, HowTo (75L) only in PROJECT. The blocking question
("does autolens_workspace_test's timeout/kill get promoted, or keep a
documented divergence?") was already answered by PROMOTION: timeout_for
and kill_group live in PyAutoHands build_util (#226/#227 -> 52408a84) and
all 10 copies now import them with guarded fallbacks. Prompt steps 1
(skip-guard) and 4 (_BUILD_DIR) verified done, not redone.
Decision: full delegation to a PyAutoHands-owned runner, staged. The
zero-drift measured is the receipt for three manual N-repo sweeps, not
evidence the copy-per-repo design is safe; the HowTo tier needed none of
them because it holds no logic. The real blocker is discovery model, not
behaviour — run_python.py is opt-out with no notebook leg, the other two
variants are opt-in allowlists — so each phase is a Hands feature first
and per-repo edits second.
- prompt draft/maintenance/ci/ -> active/, Status: issued, Issued: 2026-08-24
- active.md: smoke-runner-delegation, status library-dev,
worktree_check_conflict clean (exit 0) across all 8 repos
- dashboard regenerated
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UpSFum81Jeq9KZ9wdKtaeZ
--list allowlist mode landed on PyAutoHands claude/smoke-copy-drift-ci-docs-ozntvv (8884ecd): opt-in script lists for run_python.py, 8 tests pinning order / no_run-wins / stale-entry-continues / missing-list-errors, and the corrected run_smoke.py copy inventory in docs/internals.md. Registry now claims PyAutoHands for this task. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UpSFum81Jeq9KZ9wdKtaeZ
run.py gains --list, --no-write-back and --retry-from (c979b37), so the notebook-capable workspace variant can collapse too. Records the two phase-2 findings that changed the plan: run_notebook.py's in-place write-back is wrong for a PR gate (hence --no-write-back), and JUPYTER_MISSING_RC did not need promoting because build_util never shells out to a bare jupyter binary. Corrects the earlier progress note: with the full toolchain installed the Hands suite is 388 passed / 5 skipped / 0 failed — the 14 failures noted before were missing test tools, not pre-existing breakage. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UpSFum81Jeq9KZ9wdKtaeZ
1 task
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Mind state for the
smoke-runner-delegationtask (PyAutoHands#260, code inPyAutoHands#261).
The prompt
run_smoke_copy_drift.mdgated on one written decision before anyrepo could be touched. This records it, along with a re-measurement showing the
prompt's own 2026-08-05 correction had itself gone stale.
What changed
draft/maintenance/ci/run_smoke_copy_drift.md→active/,Status: issued,Issued: 2026-08-24, carrying the re-measurement and the decision.active.md— registerssmoke-runner-delegation,status: library-dev,claiming PyAutoHands, with both Hands legs' progress and the phase-2 findings.
dashboard.md/dashboard.html— regenerated (pyauto-brain intake --apply dashboard).The re-measurement
Measured from every repo's
main, not inferred:_workspace_testPROJECT =)Every blocker the prompt named is already closed:
timeout_forandkill_grouplive inbuild_util(record: nss-remainline-mainline-blackjax shipped (PyAutoFit#1492 + workspace_test#86) #226/prompt: record lacosmic CR dial + star-pass decoupling shipped (PyAutoReduce#66) #227 →52408a84); all ten copiesimport them with guarded fallbacks.
_kill_groupexists in zero copies.is_clean_skip_exitin exactly the threenotebook-capable copies — correct, the other seven never invoke
jupyter._BUILD_DIR) done — autogalaxy_workspace is byte-identical.The decision
Full delegation to a PyAutoHands-owned runner, staged. The zero drift above
is the receipt for three manual N-repo sweeps (#185, #226/#227, the jupyter
guard), not evidence the copy-per-repo design is safe — the HowTo tier needed
none of them because it holds no logic. The real blocker was discovery model,
so each phase is a Hands feature first and per-repo edits second.
Test Plan
lifecycle.py check→ OKlifecycle.py dates→ this task's prompt and registry row both datedTwo pre-existing undated drafts (
smoke_relevance_gate.md,slowest_smoke_gate_scripts.md) are unrelated to this task and left untouched.Generated by the PyAutoLabs agent workflow.
Generated by Claude Code