[napetrov]

Follow-up to #153 (#41 D4); answers the review question on uptake.py:109 that was still open when #153 merged.

Summary

A bare name at the start of a command counted as a run of the arm's root-level file. It resolves through PATH, and the workspace is not on PATH, so it does not run that file.

 root-level added file "lint", command text:
   ./lint x            -> counted
   /eval/run/workspace/lint -> counted
   sh lint             -> counted
-  lint x              -> counted
-  cd d && lint        -> counted
+  lint x              -> not counted
+  cd d && lint        -> not counted

Evidence

• tests/test_repo_uptake.py: lint x, cd d && lint and bare run move to the no-match cases; 40 passed. ruff clean.
• Not covered: cd d && ./lint still counts; the matcher does not follow cd (same limit as the nested-path rule).

Merge Danger

Door: two-way

Blast Radius: small. Only uptake.commands of patch: cells whose arm adds a root-level file; a count can drop.

🤖 Generated with Claude Code

[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 91.68%. Comparing base (565f653) to head (def5208).
⚠️ Report is 1 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     #186   +/-   ##
=======================================
  Coverage   91.67%   91.68%           
=======================================
  Files         223      223           
  Lines       27505    27520   +15     
  Branches     4525     4530    +5     
=======================================
+ Hits        25216    25231   +15     
  Misses       1620     1620           
  Partials      669      669           

Files with missing lines	Coverage Δ
dokimi/dokimi/repos/uptake.py	93.33% <100.00%> (ø)

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

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

[xaleryb]

• Ran: main and head each on 36 command strings through arm_uptake, the new tests against the pre-fix uptake.py, a mutation pass on the new regex, and the full suite, ruff, mypy and bandit at the head.
• Only lint and lint x change (1 -> 0). The other 34 inputs give the same result on main and head, the negatives included (./lint.sh, a/lint, ../lint). cd d && ./lint still counts, as the description says.
• The new test cases fail on the pre-fix source and pass at the head.
• The open question from feat(repo): record whether the agent used what a patch: arm added (#41 D4) #153 (lint x / cd d && lint) is answered: both are 0 at the head.
• The only failure is the Tests jobs, all in test_growth_ratchet. It comes from the base and is fixed on main by ci(growth): take #153's growth into the baseline; name the PR behind each post-#93 entry #183, so this needs a rebase or CI re-run before merge.
• Nothing is left for the author beyond that re-run.

[antonwolfy]

LGTM. The fix is right: a bare root-level name no longer counts, and ./name, the workspace path and the interpreter forms still do. One nit on the regex.

What I ran, on a Linux review host at a20a1ff:

• make ci-local: 27 passed, 2 failed, 7 skipped; 7246 tests, total coverage 93%.
• The two failures are not the PR's. test_growth_ratchet reports dokimi/repos/compare.py::_provenance_warnings: 11 complexity is over the 10 cap, and the PR does not touch compare.py. pip-audit fails on the lock files, which the PR does not touch.
• Base and head regexes compared on about 70 command strings. Only bare-name forms (lint, lint x, x | lint, (lint), cd d && lint) change from counted to not counted. Names with a / behave the same on both.
• Mutation: restoring (?:\./)? in the new pattern fails 3 of the new test cases; restoring it in only one of the two branches fails 1 or 2.

Checked, no issues

• Docs: no other doc or docstring still says "first in a command"; the contract line and the module docstring match the new behaviour.
• Changelog: the fragment is in the right directory and follows the fragment format.

[antonwolfy]

LGTM. The fix is right, and the merge with main cleared the test_growth_ratchet failure that the earlier head had. The one nit from my earlier review is still open.

What I ran, on a Linux review host at 2011ba3:

• make ci-local: 29 passed, 0 failed, 7 skipped. tests/test_repo_uptake.py passes.
• Base and head patterns compared on 73 command strings and 5 file sets. All 79 differing results go from counted to not counted, and every one is a bare root-level name as the first word of a command or of a ;, &&, | or ( part. ./lint, the workspace path, sh lint, python3 lint, x;./lint, x|./lint, $(./lint) and names with a / give the same result as the base.

Still open from my earlier review:

• dokimi/dokimi/repos/uptake.py:109: the [;&|(]\s*\./ alternative is still in the pattern. Removing it changes no result on the 73 strings, so it is dead code, and the x | lint, (lint), ;lint no-match cases are still not in the test.

Checked, no issues

• Docs: the contract line and the module and function docstrings match the new rule. No other doc still says "first word of a command".
• Changelog: the fragment follows the format in changelog.d/README.md.
• Repo rules: no schema, nav or generated-file change; the tests use synthetic events and run offline.

[antonwolfy]

LGTM. I pushed a commit per nit, up to def5208, which fix the nit from my review:

• def5208: dokimi/dokimi/repos/uptake.py:109: removed the dead [;&|(]\s*\./ alternative and added x | lint, (lint) and ;lint to the no-match test cases.

make ci-local at def5208: 29 passed, 0 failed, 7 skipped.