Conciseness

The body is mostly efficient but repeats itself noticeably: the read-before-append directive appears in Prerequisites, step 1 of When to Use, the first Anti-Pattern, and Mindset; 'When to Use' restates the description's full trigger list; the one-sentence 'Where Rules Live' section adds nothing beyond Quick Start; and 'TYPICALLY this is the same command as the Quick Start validation step' is meta-commentary. This matches anchor 3 ('mostly efficient but includes some unnecessary explanation or could be tightened') better than anchor 4, where over-explanation would be only minor.

Actionability

Fully executable, copy-paste-ready commands cover the common case: 'cat .claude/RULES.md', './scripts/validate-rules.sh validate .claude/RULES.md', and './scripts/validate-rules.sh generate "<title>" "<directive>" "<rationale>"', backed by a concrete entry template and real schema/template bundle files. This matches anchor 5.

Workflow Clarity

A clear numbered 1-5 sequence (read, dedupe-check, append via generate, inform user, validate before done) with an explicit validation checkpoint before treating the change as complete. It falls short of anchor 5 because there is no error-recovery feedback loop saying what to do when 'validate-rules.sh validate' fails (fix and re-validate), which the anchor-5 example includes.

Progressive Disclosure

The References section clearly signals existing one-level-deep files (references/rule-format.md, assets/schemas/rule-entry.schema.json, assets/templates/rule-entry.yaml) and the Quick Start orients the reader immediately. However, the body's 'Rule Format' section duplicates the core of references/rule-format.md inline rather than fully splitting it out, a minor organization gap that fits anchor 4 rather than anchor 5.

Specificity

The description names the domain ('project-level agent behavioural rules recorded in a single rules file') and several concrete actions ('Read the rules file to check existing rules before adding a new one') plus explicit negative scope ('DO NOT use for ephemeral notes, one-off instructions...'). It exceeds anchor 3 (only 1-2 actions) but falls short of anchor 5 since coverage has gaps (no amend/merge/delete of rules).

Completeness

It explicitly answers 'what' (manage/read/dedupe/append project rules in a single rules file) and 'when' ('Use when the user says...' with concrete trigger phrases), plus a when-not clause. This clearly matches the anchor-5 example structure of explicit what AND when with concrete triggers.

Trigger Term Quality

Comprehensive natural-language trigger coverage with many synonyms a user would actually say: "new rule", "add a rule", "record that", "make a rule that", "note this", "take note of", "new instruction", plus an explicit consolidated 'Triggers:' list. No common variation is missing, matching anchor 5 rather than anchor 4.

Distinctiveness Conflict Risk

The niche (a single project rules file) is clear and mostly distinct, but phrases like "note this", "take note of", and "record that" are things users also say when they want general note-taking or memory capture, creating minor overlap risk with closely related skills. The 'DO NOT use for ephemeral notes' clause only partially mitigates this, so anchor 4 fits better than anchor 5.