[antonwolfy]

Summary

Leftovers from the reviews of #66 and #109: a nested Cursor rules directory that repo inspect missed, and two comments that no longer match the code.

Changes

• Nested .cursor/rules (raised by xaleryb on #66, endorsed in my review). detect_instructions matched .cursor/rules/*.mdc only at the repository root, so web/.cursor/rules/style.mdc gave no instruction record at all. Such a file is now listed with directory web. Its scope and activation are unknown, and its note says why: Cursor's rules docs (cursor.com/docs/context/rules) cover rule subfolders inside the root .cursor/rules and nested AGENTS.md, not a .cursor/rules below the root. So whether Cursor reads one is not known from the docs; a reach probe with a Cursor profile could measure it. A path under the root .cursor/rules/ (.cursor/rules/sub/.cursor/rules/b.mdc) stays the root rule's only, and a doubly nested one takes the directory above the closest .cursor/rules/. The contract's instructions list says rules may sit at the root or in a subdirectory, and that a record of unknown scope gives the directory the file would apply to.
• ci_local.py docstring (#109 review, nit 2): it named static as the only job whose steps run under !cancelled(). The deptry step of typecheck does too.
• TRACKED_DEPS comment (same review, nit 3): it called every entry a direct dependency. ragas is in the ragas extra, anthropic comes with that extra's langchain-anthropic, and an entry that is not installed is recorded as None.
• Changelog fragment (Fixed) for the repo inspect change.
• The instruction-rule table test moves from tests/test_repo_environment.py to a new tests/test_repo_instructions.py, together with the new case. The old file is over the test length limit and may only shrink, and its baseline drops from 1,373 to 1,337 lines.

Risk

repo inspect (repo_environment.v1) gains records for nested .cursor/rules/*.mdc files, each with scope and activation unknown. So a repository with them also gets an instructions.activation-unknown finding for them, as one with root rules already does. The schema is unchanged.

Testing

On a Linux review host, at 524f1a5 (on main 506f7ba):

• make ci-local: 29 passed, 0 failed, 7 skipped (the Docker catalog smoke, test-latest, the cli-surface packaging step, conda-forge-deps without micromamba, docs deploy, the pip-audit report, semgrep). 7,382 tests passed, 1 skipped; changed-line coverage 100% of 2 lines.
• INSTRUCTION_TABLE gains sub/.cursor/rules/a.mdc. A new test pins the record's directory, its unknown scope and its note, and that a non-rule file under .cursor/ is not matched. Dropping the rule fails 2 tests; a rule without the directory fails 1.
• Rebased onto main 92d92d3 (test(llm): pin provider-call budgets for judge, panel, answer and validation paths #170, tests only) before opening; with the ten follow-up branches merged together, the full suite (7,447 passed, 1 skipped), the growth check, mypy and ruff pass.
• Review fixes, at 95ed677: make ci-local 29 passed, 0 failed, 7 skipped; 7,409 tests passed, 1 skipped; changed-line coverage 100% of 4 lines.

Checklist

• Tests added/updated and pytest passes
• ruff, mypy, and schema checks pass
• Docs updated (or N/A) — see the docs-drift checklist in CONTRIBUTING.md
• No ad-hoc run artifacts committed
• New files carry the SPDX license header (or are exempt in .github/.licenserc.yaml)

[vchamarthi]

@exai-bot /review

[vchamarthi]

@exai-bot /review

[exai-bot]

@vchamarthi Another request on this pull request is already running. Ask again when it finishes.

[exai-bot]

Note

AI-generated review of 618a61e4df69, requested by @vchamarthi, using the default review skill. It can be wrong: verify a finding before acting on it. React with a thumbs up or down on a finding to tell the bot whether it helped.
Model us.anthropic.claude-sonnet-5-5, effort default, 4 of 80 turns, 0.2 of 40 minutes, 65k in / 1k out tokens.

The change adds a rule so repo inspect records a nested <dir>/.cursor/rules/*.mdc as a path-specific instruction with directory=<dir> and activation unknown. It also moves the instruction table test to its own file and fixes two stale comments. I found no defects. The new regex needs a non-empty prefix, so the root rule and the nested rule never both match a root path. The existing activation-unknown finding and fixtures only reference the root style.mdc, so they are unaffected.

No issues found.

Not verified

I did not run the tests, the growth check or the ci-local run. I did not check that the growth_baseline.json line count (1337) matches the file after the removal.

[vchamarthi]

No blockers. 1 non-blocking comment.

No code defects found; one question on the documentation source for nested .cursor/rules, and CI is still running at the head. The new regex does not match root .cursor/rules/sub/a.mdc (no /.cursor/rules/ substring), so root rules keep a single record; the ci_local.py comment matches the deptry step's if: ${{ !cancelled() }} in ci.yml, and the manifest.py comment matches the ragas extra in pyproject.toml.

CI at send time: 0 failing, 11 pending.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.97%. Comparing base (92d92d3) to head (95ed677).
⚠️ Report is 13 commits behind head on main.
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #200      +/-   ##
==========================================
+ Coverage   91.78%   91.97%   +0.18%     
==========================================
  Files         223      223              
  Lines       27623    27713      +90     
  Branches     4549     4553       +4     
==========================================
+ Hits        25354    25488     +134     
+ Misses       1606     1575      -31     
+ Partials      663      650      -13     

Files with missing lines	Coverage Δ
dokimi/dokimi/repos/environment/instructions.py	97.29% <100.00%> (+0.15%)	⬆️
dokimi/dokimi/runner/manifest.py	89.28% <ø> (ø)

... and 15 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[exai-bot]

Note

AI-generated review of 94a88a16cee5, requested by @vchamarthi, using the default review skill. It can be wrong: verify a finding before acting on it. React with a thumbs up or down on a finding to tell the bot whether it helped.
Model us.anthropic.claude-sonnet-5-5, effort default, 7 of 94 turns, 0.6 of 48 minutes, 124k in / 4k out tokens.

Re-review of 94a88a16cee5: 2 new findings.

2 findings are on the changed lines, as comments below.

[napetrov]

Read the diff: nested .cursor/rules/*.mdc is recorded with scope/activation unknown and an honest note, the root rule is unaffected, the table test covers sub/.cursor/rules/a.mdc, and the comment fixes match the code. LGTM.