Repository navigation
perf(pr-management-code-review): select, scan and dispose as code, one file per outcome - #1585
Merged
Merged
Conversation
potiuk
added
family:pr-management
github-actions
Bot
added
family:setup
potiuk
force-pushed
the
perf/pr-mgmt-8-code-review
branch
from
October 10, 2026 09:46
780dfb6 to
73e30c8
Compare
potiuk
force-pushed
the
perf/pr-mgmt-8-code-review
branch
from
October 10, 2026 10:18
73e30c8 to
63ad744
Compare
…e file per outcome The code-review skill loaded ~38k tokens per run and had the agent compute, by hand, the review queue (the five "my reviews" signals and every selector), CODEOWNERS matching, the slop signals and threshold, the security and AI-disclosure scans, licence and artifact checks, the dependency-constraint ledger, the disposition, reviewer suggestions, inline-comment anchoring and the review payload. All of that now runs in `pr-management code-review …` over reads saved with `vetted-op-read --save`; the agent reads the diff, writes the findings and judges severity. A run loads ~13.6k tokens, plus the classification and criteria docs a PR actually needs. Fixes found on the way: "touching mine" uses GitHub data, not `git log --author` against a login; CODEOWNERS team ownership checks team membership, not org membership; the three contradictory APPROVE definitions collapse to one (nothing above a nit, real CI green, no unresolved threads or standing change request, no open question); `maintain` permission gets the maintainer footer. Each selector, finding category, verdict, slop outcome, footer variant and skip reason has its own document under `classifications/` or `criteria/`. 13 model-graded suites (70 cases) become pytest cases named after each case; the 15 judgement suites stay model evals. Generated-by: Claude Opus 5
potiuk
force-pushed
the
perf/pr-mgmt-8-code-review
branch
from
October 10, 2026 11:16
63ad744 to
2b05020
Compare
This was referenced Oct 12, 2026
11 of 18 tasks
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
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
pr-management code-review …over reads saved withvetted-op-read --save. The agent reads the diff, writes the findings and judges severity. A run loads ~13.6k tokens, plus only the classification and criteria docs a PR actually needs.classifications/orcriteria/.Fixes found on the way:
git log --authoragainst a login.maintainpermission gets the maintainer footer.Type of change
Test plan
prek run --all-filespassestools/pr-managementpytest / ruff / mypyreview-handoff/case-3-re-draft-edit, passed on two re-runs of the suite and onmain, so it is nondeterministic, not a regression.Was generative AI tooling used to co-author this PR?
Generated-by:.🤖 Generated with Claude Code
https://claude.ai/code/session_01GTPsZ5aEE5bUVYqp47Dpvv