bib-parse-compat

name	bib-parse-compat
description	Use when the user wants to check or analyze badness's BibTeX parse concordance against texlab — "run bib-parse-compat", "bib parse concordance vs texlab", "analyze bib divergences", or after a `.bib` parser/CST change. Runs the soft gauge and triages any unexplained divergence.

bib-parse-compat

A soft differential gauge of badness's BibTeX CST against texlab's BibTeX CST over crates/badness-parser/tests/bib_corpus/*.bib. It is not a quality gate: per AGENTS.md we measure against texlab, never match it. The skeleton stops at entry types and field names; cite keys and value internals are intentionally dropped because the two legitimately diverge there. Note texlab's bib parser has no error channel, so this gauge never reports texlab parse errors.

Run it

task bib-parse-compat

This runs cargo test -p badness-parser --test bib_parse_compat -- --ignored --nocapture and rewrites the report at .agents/skills/bib-parse-compat/BIB_PARSE_COMPAT.md (the generated artifact next to this skill — do not hand-edit it).

Analyze

1. Read the regenerated BIB_PARSE_COMPAT.md. The gauge currently sits at 100% concordance with 0 unexplained divergences, so the usual job is "confirm it is still clean."
2. If "Unexplained divergences" is 0, report that and stop.
3. For each unexplained divergence, classify it against crates/badness-parser/tests/bib_parse_compat_allowlist.toml:
   • Deliberate deviation (badness is the faithful reading): add a [deviations] entry keyed by the corpus filename with a one-line reason; re-run to confirm.
   • Genuine parser modeling gap: fix it in the bib parser per tenet 3, with corpus + snapshot tests and a losslessness assertion. Do not allowlist it.