A Run's outcome says whether its Engine step started, and the run record carries it (#757's first part, record 0201)
User description
Refs #757 ("A Run that ended environment does not use up the item's run ceiling"): its first part, by the Factory Builder's ruling of 2026-11-12. It does not close the issue; the counting is the next pull request, rebuilt from #769 on this one.
This changes harness/ and .github/workflows/, a model Lens's inputs: the Maintainer approves it by hand as donpetry-bot, and the Factory never approves it. It is held factory:hands-off.
What changes
| The run job | Hands on its Engine step's outcome as the output engine, ${{ steps.engine.outcome }}. GitHub skips the step once a step before it failed, as the CLI install did in the runner DNS outage |
apply (harness/apply_step.py) | Writes outcome/1's engine_started: false where the output is skipped, true where it is success, failure or cancelled, left out otherwise |
outcome/1, run-record | An optional boolean engine_started |
| The records port | Reads it with a failed Run's outcome into the run record, and keeps it with a kept outcome from the last two pass records (records.engines, beside records.begun) |
| The pass's entry | Hands engines to the records port |
tools/check_workflows.py | A job may hand on a step's outcome, one of four words GitHub gives, beside a step's own output; its copy of run-jobs.yml carries the new output |
| What reads it | Nothing yet |
It fails closed: the field says false only where GitHub said the Engine step was skipped. A missing output, a job killed before it handed on its outputs, an unknown word, or an outcome from before this change leaves it out, and the next change will count such a Run as one that may have reached the model.
The record is docs/decisions/0201-…. 0201 is the Factory Builder's number, behind 0197 (#774), 0198 (#761), 0199 (#760) and 0200. Until those merge, the checks legs are red on "records are numbered from 0001 with no gap", and nothing else fails. The number is confirmed before the final push.
Checks
- Boundary tests, written first and watched failing:
tests/test_steps.py::test_apply_says_whether_the_engine_step_started(each word, the live workflow's output andNEEDS),tests/test_adapters.py::test_whether_the_engine_started_is_read_with_the_outcome_and_kept_by_its_run_and_attempt, the outcome's own read intest_a_runs_outcome_is_read_from_its_one_file_and_held_to_outcome_1, the entry intests/test_controller.py, and the expression rule intests/test_tools.py. - 22 new mutants and 6 retargeted, each killed.
make checksis clean but for the provisional number. The desk's unfiltered run at a46e5c7a, the record renumbered in a copy to clear the gap: 6,900 of 6,900 mutants killed. After main's merge, every mutant ofcontroller/__main__.pyandadapters/records.pywas run again: 205 of 205 killed. - Not verified on GitHub: that a job's output of
steps.engine.outcomereadsskippedafter an earlier step failed. GitHub's documentation givessteps.<id>.outcomeas one of success, failure, cancelled or skipped. No trial has run, since a session may not dispatch a workflow. If it reads otherwise, the field is left out, and the Run counts as it does today. - About 27,600 changed characters.
Protected paths
.github/workflows/, harness/, adapters/, controller/, schemas/, tests/ and tools/ are protected or reported paths: pre-approved by the Maintainer 2026-11-08, ruled by the Factory Builder (its ruling of 2026-11-12 on #757). harness/ and the workflows are model Lens inputs, so the Maintainer approves this pull request by hand.
Reviews
| Reader | Head | Findings | Disposition |
|---|---|---|---|
| Claude security pass | b02ece00 | 0 | clean |
| Claude review | b02ece00 | 2 minor, 3 info | Fixed in a46e5c7a: the entry's reading from records before the pass is now asserted (its mutant restored); a counter-example holds the run record's engine_started to a boolean; the outcome read's guard has a test and a mutant; engines() reads the last record alone, since known keeps a fault from it alone and an earlier word could never reach a run record. Rollback: an older Release reads a newer outcome as no outcome/1, named as a cost in record 0201, as began_at already is |
| Carbon (gemini-4-carbon-high) | b02ece00 | 3 minor, four parts read | Fixed in dad97328: "failure" and "success" each have a mutant of its own. The previous[-2:] bound is gone, since engines() reads the last record alone (a46e5c7a). The apply job's NEEDS: ${{ toJSON(needs) }} is held by the new test's read of the live workflow, and by every apply test, since without it nothing is read |
| Muse (muse-wonder-full-1.0) | a46e5c7a | 1 major | The numbering gap: by design. The record's number is the Factory Builder's, now 0201, behind 0197 to 0200 |
| Astra (gpt-6-astra) | dad97328 | 1 major | Fixed in ed1a1f64: a kept outcome is tested against what was said of another Run and of another attempt, with a mutant that takes any Engine word kept |
🤖 Generated with Claude Code
CodeAnt-AI Description
Record whether a run reached the Engine
What Changed
- Failed-run records now say whether the Engine step started: false when it was skipped, and true when it succeeded, failed, or was cancelled.
- If the step’s status is missing or unrecognized, the record leaves it unset rather than claiming the Engine did not start.
- The status is retained when a failed run’s outcome is carried into later records.
Impact
✅ Identify runs that never reached the model
✅ Preserve Engine-start status across records
✅ Avoid false “not started” results when status is unavailable
💡 Usage Guide
Checking Your Pull Request
Every time you make a pull request, our system automatically looks through it. We check for security issues, mistakes in how you're setting up your infrastructure, and common code problems. We do this to make sure your changes are solid and won't cause any trouble later.
Talking to CodeAnt AI
Got a question or need a hand with something in your pull request? You can easily get in touch with CodeAnt AI right here. Just type the following in a comment on your pull request, and replace "Your question here" with whatever you want to ask:
@codeant-ai ask: Your question here
This lets you have a chat with CodeAnt AI about your pull request, making it easier to understand and improve your code.
Example
@codeant-ai ask: Can you suggest a safer alternative to storing this secret?
Preserve Org Learnings with CodeAnt
You can record team preferences so CodeAnt AI applies them in future reviews. Reply directly to the specific CodeAnt AI suggestion (in the same thread) and replace "Your feedback here" with your input:
@codeant-ai: Your feedback here
This helps CodeAnt AI learn and adapt to your team's coding style and standards.
Example
@codeant-ai: Do not flag unused imports.
Retrigger review
Ask CodeAnt AI to review the PR again, by typing:
@codeant-ai: review
Check Your Repository Health
To analyze the health of your code repository, visit our dashboard at https://app.codeant.ai. This tool helps you identify potential issues and areas for improvement in your codebase, ensuring your repository maintains high standards of code health.