Skip to content

chore(v1.10): rules-audit fixes - author_title items 1-8 - #73

Merged
jameskhair-code merged 1 commit into
mainfrom
chore/v1.10-rules-audit
Jun 11, 2026
Merged

jameskhair-code merged 1 commit into
mainfrom
chore/v1.10-rules-audit

Conversation

@jameskhair-code

Copy link
Copy Markdown
Owner

Item 1 of the v1.10 preflight (docs/planning/v1.10-charter.md). Prompt-content only — no code changes.

What changed

The eight rules-content findings from the 2026-06-10 title/author audit, in priority order:

  1. T-SER-02 contradictory example (rules/author_title.md): the WRONG line showed the correct transformation, byte-identical to the RIGHT line. It now shows the actual failure mode (leaving the series parenthetical unchanged).
  2. T-SUB-07 duplicated alternative: separated by " / " (or " / ") — both strings identical; now (or "/" without spaces).
  3. GEN-01 vs output format: aligned the no-change note to "Already correctly formatted." — the string the output format mandates. ai/authors.py:128 tolerates both strings case-insensitively, so this is prompt-only.
  4. GEN-07 notes style: examples harmonized on rule-ID citations ("Removed marketing suffix per T-AWD-04") for review/calibration traceability.
  5. A-MUL bridge note: the section's separator rules describe parsing the stored single-string field; a new note maps that onto the JSON authors-array output — one person per element, no "&" inside an element except single corporate names (A-ROL-11 / A-SPE-07).
  6. A-MUL-05: the flag now triggers on "et al." alone, not "more than three authors AND et al."
  7. T-CAP-10: the stylistic-lowercase example was an author (bell hooks); replaced with a real lowercase title ("milk and honey", Rupi Kaur).
  8. confidence.md SECTION CONF index: now lists author_title.md with a pointer to the inline GEN-02..GEN-05 guidance.

Audit item 9 (normalize.py typographic quotes) stays parked on the roadmap pull-list pending the mapping sign-off.

Verification

  • python -m pytest -q — 639 passed.
  • Real-library smoke (per the charter's gate for this item): two clean-titles --dry-run batches of 8 against the live library.
    • Batch 1 (not #mqg_title_author:true): 8/8 already clean; no-change responses parsed cleanly.
    • Batch 2 (titles with parens/colons): 6 changes suggested, 2 already clean. All six proposals correct (Tier-A subtitle removal, author-prefix strip, T-FMT-09 dash fix, 3x T-SER-01 volume standardization), and the notes came back in the new rule-ID-citation style. ~$0.02, no writes.

🤖 Generated with Claude Code

Eight prompt-content fixes from the 2026-06-10 audit:
1. T-SER-02: WRONG example showed the correct transformation; now shows
   the left-unchanged failure mode.
2. T-SUB-07: duplicated alternative in the separator parenthetical; now
   reads 'or "/" without spaces'.
3. GEN-01 aligned to the output format's "Already correctly formatted."
4. GEN-07 examples harmonized on rule-ID citations.
5. A-MUL section: bridge note mapping the single-string parse rules onto
   the JSON authors-array output shape.
6. A-MUL-05: flag triggers on "et al." alone.
7. T-CAP-10: stylistic-lowercase example is now a title (milk and honey),
   not an author.
8. confidence.md SECTION CONF index now lists author_title.md (GEN-02..05).

Code-side tolerance for both no-change note strings confirmed at
ai/authors.py:128. Item 9 (normalize.py quote mapping) stays parked per
the roadmap pull-list.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
@jameskhair-code
jameskhair-code merged commit 94a2067 into main Jun 11, 2026
2 checks passed
@jameskhair-code
jameskhair-code deleted the chore/v1.10-rules-audit branch June 11, 2026 01:42
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant