[antonwolfy]

Summary

Follow-up to four merged PRs whose reviews named checks that no test held. On main, each of the lines
below can be changed in one place (the mutant in the table) and the whole suite still passes; the code is
right, the tests did not notice. This PR adds one test or parametrized case per point and changes no code.

• feat(context): prefilter AGENTS.md rules against session logs (candidates for ablation) #334 (context prefilter), review:
  the redirect target's < exclusion, the exact-name match and the \ handling in _file_named, and
  strip_wrappers in _prints. No author reply.
• feat(repo): repo budget counts .claude/rules #326 (repo budget), review
  (null / ~ in _names_pattern, no author reply) and
  thread (the docs/alias.md variant
  the reply left out, and the RULE_GLOB.match(f) half of _counted_import).
• repo skill-reach: measure which Agent Skills a harness lists #363 (repo skill-reach), thread:
  the reply covered --timeout, --salt and --budget. This PR adds the run path's OSError clause,
  --bin and --keep.
• feat(repo): dokimi_config.v1 schema and repo_config loader #323 (repo config), review B1
  (**/Makefile against a root file) and
  review B2 (glob.escape on the
  config's link hops, the .. prefix and the commonpath guard in link_text, --literal-pathspecs). The
  author left B2 as is because it named no specific tests; this PR names them.

Point	Code on main	Mutant that passed the suite	New case (passes on main, fails on the mutant)
#334	prefilter.py _REDIRECT	drop < from the write target's class	cat >notes.txt<<docs/guide.md …: acted 0 (the mutant gives 1)
#334	prefilter.py _file_named	key in name	an Edit of /repo/Makefile.bak or /repo/GNUMakefile is no action for rule Makefile
#334	prefilter.py _file_named	drop _slashes(...)	an Edit of C:\r\Makefile acts on rule Makefile
#334	prefilter.py _prints	drop strip_wrappers	sudo echo changelog.d/x.md: acted 0
#326	rule_frontmatter.py _names_pattern	if part == "":	paths: [null and paths: [~ (the key does not parse) load at start
#326	budget.py _import_text	drop the RULE_GLOB.match(_target(...)) half	a rule reached through docs/alias.md does not import the @../extra.md in its frontmatter
#326	budget.py _counted_import	reduce to RULE_GLOB.match(t)	@.claude/rules/r.md, a link to a plain shared.md: the rule entry keeps its 5-char body count (the mutant makes it an 18-char import)
#363	repo_skill_reach.py run handler	drop OSError	a directory at out/skill-discovery.json: exit 1 and a ✗ message, no traceback
#363	repo_skill_reach.py	bin_path=None	--bin naming a wrapper that runs the fake harness in crash mode: the run reports its exit 2
#363	repo_skill_reach.py	keep=False	--keep: the fixture stays under --work-dir and the artifact's workspace names it
#323	repo_claims.py _match	states = {0}	shared: ['**/Makefile'] claims a changed root Makefile
#323	repo_config.py _own_path	drop glob.escape on the hops	a change to the link c[1] on the way to the config is shared, c1 is not
#323	utils.py link_text	"/".join(rest)	sub/l -> <root>/../t/x reads as ../x, not x
#323	utils.py link_text	key.startswith(key_anchor) alone	root <tmp>/t, link to <tmp>/t2/q.json: kept absolute, not 2/q.json
#323	utils.py _tracked_modes	drop --literal-pathspecs	a tracked :x.sh takes its mode from the index, not the disk

The #334 review gave echo hi >out.txt<docs/guide.md as the case, with acted False on main. At this
head it gives acted 1 on main and on the mutant alike: <docs/guide.md is a read redirection, which the
prefilter counts as touching the file. A heredoc sign glued to the write target (>notes.txt<<docs/guide.md)
is where the two differ, so that case pins the line.

Changes

• tests/test_context_prefilter.py: a sudo echo case for the printed-path test, and
  test_a_bare_file_name_claim_counts_an_edit_of_that_name_only_on_any_separator.
• tests/test_context_prefilter_shell.py: a glued heredoc case for
  test_a_heredoc_delimiter_word_is_not_a_path_the_command_touched.
• tests/test_repo_budget_rules.py: three unparsed-key cases for
  test_only_a_paths_value_that_names_a_pattern_scopes_a_rule, and two symlink tests next to
  test_an_alias_of_a_rule_is_followed_as_an_import.
• tests/test_skill_reach.py: three CLI run tests (artifact save OSError, --bin, --keep).
• tests/test_repo_config_globs.py, tests/test_repo_config_links.py, tests/test_tree_digest.py: one test
  for each feat(repo): dokimi_config.v1 schema and repo_config loader #323 point.
• No changelog fragment: nothing under dokimi/dokimi/ changes.

Risk

None: tests only, and the output is unchanged. The symlink tests skip where symlink_to raises. The :x.sh
test skips on Windows and without git. The --bin wrapper test skips on Windows (it is a POSIX shell
script).

Testing

On a Linux review host, at 9e07365 (on main 51bbc4d):

• make ci-local: 29 passed, 1 failed, 7 skipped. Skipped: external-catalog-smoke (needs docker),
  test-latest, docs.yml:deploy and pip-audit.yml:report (job conditions false on a PR), the
  cli-surface "Package the result" step, conda-forge-deps (no micromamba) and semgrep (offline).
  15610 tests passed, 9 skipped; changed-line coverage has no measurable lines (tests only).
  The failed step is agent-quality.yml:benchmark / Run benchmark, and it fails on main too:
  golden_manifest violation: matched 40 document(s) ... more than max_docs=39. feat(repo): rule ledger: repo compare --ledger with per-rule keep/review/insufficient-evidence verdicts #336 added
  docs/rule-ledger.md without raising max_docs in benchmarks/spec.v1.yaml. This PR adds no doc and
  leaves that alone.
• The seven touched test files with test_docs_code_refs.py, test_growth_ratchet.py,
  test_audit_test_quality.py and test_import_architecture.py: 851 passed, 2 skipped (existing tests).
  ruff check tests is clean.
• Mutants: each of the 15 changes in the table, applied to a copy of the committed tree, makes its new
  test fail; the same tests pass on the unchanged code.
• Merged together with the other follow-up branch (fix/tree-digest-nested-ignores) on main 51bbc4d: pytest -q -n auto 15614 passed, 9 skipped; mypy and ruff check . clean.

Revert

git revert of the one commit; it removes tests only.

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
• Self-reviewed per "Before opening a PR" in AGENTS.md
• No ad-hoc run artifacts committed
• New files carry the SPDX license header (or are exempt in .github/.licenserc.yaml)

[Copilot]

🟢 Approval recommended

No active review issues block approval of this tests-only regression-coverage change.

No open issues 🎉

What changed in this PR

0 new findings; nothing to fix.

This tests-only PR adds regression coverage requested in reviews of #323, #326, #334 and #363, protecting existing dokimi behavior without changing production code.

Changes:

• Covers context-prefilter filename matching, command wrappers and heredoc handling.
• Pins rule-budget behavior for malformed frontmatter and symlinked imports.
• Adds skill-reach CLI error and option checks, plus config-path and tree-digest edge cases.

File	Description
dokimi/​tests/​test_tree_digest.py	Checks symlink normalization and index modes for colon-prefixed filenames.
dokimi/​tests/​test_skill_reach.py	Tests artifact-write failures, --bin forwarding and --keep.
dokimi/​tests/​test_repo_config_links.py	Checks literal matching of bracketed config-link paths.
dokimi/​tests/​test_repo_config_globs.py	Verifies leading **/ matches root files.
dokimi/​tests/​test_repo_budget_rules.py	Covers malformed paths values and symlinked-rule counting.
dokimi/​tests/​test_context_prefilter.py	Tests wrapped printing commands and exact filenames across separators.
dokimi/​tests/​test_context_prefilter_shell.py	Checks that glued heredoc delimiters are not counted as touched paths.

_{Review effort: Balanced}

[exai-bot]

✅ OpenCodeReview: Review complete: 0 finding(s) across 7 selected item(s).

Review details

• Commit reviewed: 2a65a40 against merge base 5ce3207
• Coverage: 7 files in 4 groups
• Engine: OpenCodeReview v1.13.3, us.openai.gpt-6.1-sol
• Usage: 187k tokens (130k from cache), 17s
• Tool calls: 12 (11 file_read, 1 code_search)
• Context prefilter tests: 2 files
• Repository configuration and budget rules tests: 3 files
• Skill reach tests: 1 file
• Tree digest tests: 1 file

[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 94.76%. Comparing base (5ce3207) to head (2a65a40).
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #389   +/-   ##
=======================================
  Coverage   94.75%   94.76%           
=======================================
  Files         352      352           
  Lines       46498    46498           
  Branches     7969     7969           
=======================================
+ Hits        44061    44063    +2     
+ Misses       1544     1543    -1     
+ Partials      893      892    -1     

see 3 files with indirect coverage changes

🚀 New features to boost your workflow:

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