review agent's parser failed on collapsed generated-file markers

TL;DR: Generated files do not always come with hunks - the diff can show a collapsed generated-file marker instead, which a hunk-expecting parser chokes on. Detect the generated marker (in the diff header or the file's linguist attributes), mark the file as generated, and skip hunk parsing for it. Generated files get a file-level note, not line comments.

review agent's parser failed on collapsed generated-file markers

1. Capture the failing section. Isolate the diff chunk where the parser breaks and confirm it is a generated-file marker (a collapsed notice where hunks would normally be) rather than malformed input.

Expected: the break point sits exactly on the generated marker.

1. Add generated-file detection before hunk parsing. Check the diff header and the file's generated attributes; when the marker is present, record the file as generated with an empty hunk list and move to the next file.

Expected: the parser walks past generated sections without error.

1. Make the hunk loop defensive anyway. If a file section contains neither hunks nor a recognized marker, skip it with a warning instead of throwing.

Expected: unknown collapsed formats degrade to warnings, never crashes.

1. Route generated files around line review. The reviewer should emit one file-level note ("generated file, not line-reviewed") and never attempt line comments on it.

Expected: no comment calls reference lines inside generated content.

1. Keep an explicit generated-path list as a second line of defense (lockfiles, compiled output, vendored code, minified bundles) for diffs where the marker is absent.

Expected: generated files are excluded even when the diff does not label them.

1. Re-run the parser over the original failing PR diff.

Expected: the full diff parses, generated files are flagged, and real source files parse exactly as before.

Use this when

• The parser breaks on collapsed generated-file sections
• Generated files appear in the diff without hunks
• The agent tries to line-comment inside generated content
• You need generated files excluded from line review

Not for this skill when

• The parser breaks on binary patch sections (different marker, different fix)
• The parser breaks on renamed-file hunks (different header, different fix)
• The diff is truncated by the API (that is a fetch/pagination problem)
• A generated file genuinely needs review (review its source or generator config, not the output)

Variant phrasings

• "diff parser fails on generated file marker"
• "collapsed generated files break review parser"
• "skip generated files when parsing PR diff"
• "parser chokes on linguist-generated sections"

Why it happens

Diff UIs collapse generated files to keep PRs readable, and the underlying diff reflects that: instead of hunks you get a marker saying the file is generated and collapsed. Parsers written against full text diffs assume hunks always follow a file header, so the marker is an unexpected shape that breaks the state machine.

Edge cases

• The marker wording varies between git hosts and diff renderers. Match on the concept (generated + collapsed) rather than one exact string.
• A file can be generated and still hand-edited in the same PR. If hunks ARE present alongside the marker, parse the hunks normally and note the generated status.
• Excluding generated files from review is a policy choice. For lockfiles some teams want a dependency-change summary instead of silence - emit that at file level.
• Generated-file detection via path patterns can false-positive on hand-written files with similar names. Prefer the diff's own marker when it exists.

Provenance

Resolved from the public thread: https://vectle.com/posts/pstp0DUSJdrBCy03ADXgbpGA