[antonwolfy]

Summary

Follow-up to the review of #459, where two points were left open: a delimited URL is cut at | and at a space, and no test pins the bracket counter (T11.2, T13.2).

On main, extract_urls("<https://acme.dev/a|b>") gives ['https://acme.dev/a'] and extract_urls("[a](<https://acme.dev/x y>)") gives ['https://acme.dev/x'], although the punctuation of an autolink or a <...> destination is part of the URL, and the probe requests the wrong page. Replacing brackets -= 1 with pass in _clean_url passes the whole suite.

Changes

• dokimi/dokimi/external/online_checks.py: in _clean_url, a delimited URL (an autolink or a <...> destination) ends at its > only, so a | inside it stays. The undelimited case still ends at |.
• Table rows: a GFM table splits its row at an unescaped | before it reads links, and extraction sees the row as one text. A line that starts with | is therefore passed as table_row, and there an unescaped pipe still ends a delimited URL (| <https://acme.dev/a|b> | x | gives https://acme.dev/a). A table without a leading pipe is not detected.
• _distinct_urls reads a <...> link destination up to its >, blanks included (a trailing one too); [a](<https://acme.dev/x y>) is requested as https://acme.dev/x%20y (httpx encodes the space). An autolink holds no space under CommonMark and is unchanged. A backslash-escaped > or < (any backslash escape of ASCII punctuation) does not end or stop the scan: [a](<https://acme.dev/x y\>z>) is requested as https://acme.dev/x%20y%3Ez.
• Tests in dokimi/tests/test_catalog_online.py: the pipe in an autolink and in a destination, the table-row cases, the spaced destination with a mocked transport that sees /x%20y only, and [see https://acme.dev/a[b]c] for more, which pins brackets -= 1.
• Changelog fragment (Fixed).

Risk

A URL in <...> that holds a |, or a <...> destination that holds a space or an unmatched )/], is now probed whole instead of cut. A skill that has such a link gets its real target checked, so a link-dead can appear or go. Nothing else changes: the docs do not say where a URL ends, so none changed.

Testing

On a Linux review host, at 213e757 (on main b4bcfa4):

• make ci-local: 33 passed, 0 failed, 14 skipped (job conditions, no semgrep or micromamba, other-Python legs). 20402 tests passed, 19 skipped; the coverage of changed lines step passed.
• Repro: main prints ['https://acme.dev/a'] ['https://acme.dev/x']; this branch gives ['https://acme.dev/a|b'] ['https://acme.dev/x y'].
• The new tests on main's code: the pipe test and the spaced-destination test fail.
• Mutants, each failing the test that pins it: brackets -= 1 → pass fails test_an_unbalanced_closing_bracket_ends_the_url_after_a_balanced_pair; dropping table_row from the pipe condition fails test_a_pipe_in_an_angle_delimited_url_is_part_of_it_outside_a_table_row; not reading the angle destination to its > fails test_an_angle_destination_keeps_its_spaces_and_is_probed_with_the_path_encoded. The bracket test passes on main: it pins behaviour that already held.
• A trailing blank inside <...> is kept: extract_urls("[a](<https://acme.dev/x >)") gives ['https://acme.dev/x '] and the probe requests /x%20; with .rstrip() restored the spaced-destination test fails. An unclosed [a](<https://acme.dev/x y) falls back to the bare URL https://acme.dev/x, which covers the branch codecov reported as partial.
• An escaped bracket in a <...> destination: extract_urls gives ['https://acme.dev/x y>z'] for [a](<https://acme.dev/x y\>z>) and ['https://acme.dev/x y<z'] for the \< form, and the mocked transport sees /x%20y%3Ez only. With the previous scan the same test fails (/x y\ was read).
• Not run: Python 3.11, 3.12 and 3.14 (the host runs one version); semgrep.

Revert

A plain revert restores the cut at | and at a space. Nothing is stored.

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)

[exai-bot]

✅ OpenCodeReview: Review complete: 0 finding(s) across 2 selected item(s).

CI at 7777982: 25 passing, 3 skipped.

Review details

• Commit reviewed: 7777982 against merge base 213e757
• Coverage: 2 files in 1 groups
• Engine: OpenCodeReview v1.13.5-exai.4.8f73eada, us.openai.gpt-6.1-sol
• Usage: 50k tokens (29k from cache), 16s
• Tool calls: 4 (3 file_read, 1 local_check_results)
• small change set: 2 files

[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 95.72%. Comparing base (b4bcfa4) to head (7777982).
⚠️ Report is 8 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     #752   +/-   ##
=======================================
  Coverage   95.71%   95.72%           
=======================================
  Files         414      415    +1     
  Lines       55310    55369   +59     
  Branches     9509     9518    +9     
=======================================
+ Hits        52941    53000   +59     
  Misses       1466     1466           
  Partials      903      903           

Files with missing lines	Coverage Δ
dokimi/dokimi/external/online_checks.py	96.36% <100.00%> (+0.05%)	⬆️

... and 7 files with indirect coverage changes

🚀 New features to boost your workflow:

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

[vchamarthi]

Pushed 1 commit to fix/url-extraction-delimited (213e757..7777982).

• The template guard no longer skips a <...> destination that the angle scan closed at its >.
  Run: the full suite with main merged in, 20489 passed.