[EquityBot]

Summary

Adds a clause-dash advisory and new stock phrases to laconic-check, a calibration mode for the blocker judge, 10 frozen holdout cases and 3 skill-text additions.

laconic-check lint
├── clause-dashes (new, advisory): clause dashes in 2 or more sentences of one paragraph
│   └── counted once per sentence, so a pair around one aside counts once
├── stock phrases: + innovative, practical solutions, timeless openers
└── exit code: advisory findings never fail the run
scripts/judge.py --calibrate evals/judge-calibration.jsonl   (agreement on the blocker field)
evals/holdout-cases.jsonl   (10 frozen cases, pinned by a digest test)

Skill text: a fiction line in 'Write literally', a social post venue with a regression row, and a limits sentence in the README. Each is mirrored into .cursor/skills/.

Evidence

• Calibration: on the Style Manual copy, counting raw dashes per paragraph gave 34 findings, mostly paired asides; counting per sentence gives 2, now pinned as an advisory ceiling beside the unchanged 82-finding ceiling for other rules.
• Judge: judge.py --calibrate agreed with all 10 labelled responses (5 blockers).
• Checks: cargo test (56), clippy -D warnings, cargo fmt --check, Python unittest (168), run_evals.py validate, claude plugin validate . and git diff --check pass after rebasing on 24ce59c.

Merge danger

Door: two-way. Blast radius: lint output

The new rule is advisory, so lint exit codes do not change; output gains one finding type.

Unverified

check_context_compat.ts did not run (bun not installed locally); this change does not touch its inputs.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues
0 New dependency risks

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[codacy-production]

Up to standards ✅

🟢 Issues 0 issues

Results:
0 new issues

View in Codacy

🟢 Metrics 36 complexity · 0 duplication

Metric	Results
Complexity	36
Duplication	0

View in Codacy

AI Reviewer: first review requested successfully. AI can make mistakes. Always validate suggestions.

_{TIP This summary will be updated as you push new changes.}

[codacy-production]

Pull Request Overview

⚠️ Resolve issues before merging

The calibration workflow can still abort unexpectedly when the judge returns no rows, because it indexes the first result without handling an empty response. This should be addressed before merging.

The advisory-only CLI exit-code contract and mirrored skill consistency lack automated tests, leaving two important acceptance areas unverified. The judging entry point also has elevated complexity, which increases maintenance risk but is not itself blocking.

About this PR

• Add automated coverage for the advisory-only CLI exit-code contract and for consistency between the canonical and mirrored skill files.

1 comment outside of the diff

skills/laconic-prose/SKILL.md

_{line 1 🟡 MEDIUM RISK}
Suggestion: Add the required skill metadata, including an author, to the canonical skill file and synchronize the .cursor/skills/ mirror so the metadata is identical in both copies.

Test suggestions

• Detect clause dashes across two sentences, including em dashes, en dashes, spaced hyphens and double hyphens, while ignoring paired asides, ranges, compounds, list markers, tables and code.
• Confirm clause-dash findings are advisory and do not produce a failing lint exit code.
• Detect the added stock phrases and avoid false positives for legitimate words, protected text and unrelated forms.
• Run calibration fixtures independently, calculate agreement, report mismatches and reject an empty fixture set.
• Validate the ten holdout cases and detect any modification through the pinned digest.
• Verify the new skill guidance is synchronised between canonical and mirrored files.
• Enforce the Style Manual corpus ceilings for normal findings and clause-dash advisories.

Prompt proposal for missing tests

Consider implementing these tests if applicable:
1. Confirm clause-dash findings are advisory and do not produce a failing lint exit code.
2. Verify the new skill guidance is synchronised between canonical and mirrored files.

_{TIP Improve review quality by adding custom instructions
TIP How was this review? Give us feedback}

[codacy-production]

_{🟡 MEDIUM RISK}

Handle an empty result explicitly before indexing rows[0], either by recording a calibration error or raising a descriptive ValueError.

_{TIP React 👍 if this review is right, or reply with what's wrong and how it could be better to help improve the reviewer.}

[codacy-production]

_{🟡 MEDIUM RISK}

Suggestion: Consider extracting runner setup and calibration handling into focused helpers so main primarily parses arguments, dispatches the selected mode, and reports errors.

See Issue in Codacy
See Complexity in Codacy

_{TIP React 👍 if this review is right, or reply with what's wrong and how it could be better to help improve the reviewer.}