Content
60%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a well-organized, largely concrete reference with a clearly sequenced three-phase workflow and explicit checkpoints, but it is let down by material redundancy across the three token-budget sections and — most seriously — a progressive-disclosure architecture that is claimed but not shipped: nine of its module references are broken within the provided bundle.
Suggestions
Ship the referenced modules/ directory (or prune the pointers): every 'Detailed Reference: modules/*.md' link currently resolves to a nonexistent file, breaking the skill's own navigation model.
Consolidate the three separate token-budget treatments (sections 2 and 4 plus the closing 'Token Budget' section) into a single phase-budget table with the /clear strategy stated once.
Remove the unverifiable improvement percentages ('40 percent task success', '70 percent fewer escalations', '2-3x larger projects') and the trailing rationale sentences on each TRUST pillar — they pad tokens without adding executable guidance.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly dense, useful reference material, but includes unnecessary explanation and clear redundancy: phase token budgets (30K/180K/40K/250K) are stated three times (sections 2, 4, and the closing 'Token Budget' table), /clear guidance is repeated in sections 2, 4, and the closing section, and each TRUST pillar closes with a rationale sentence Claude does not need (e.g. 'Clear naming improves code comprehension, team collaboration, and maintenance velocity'). This fits anchor 3 ('mostly efficient but includes some unnecessary explanation or could be tightened') rather than 2, since padding is confined to duplicated token sections rather than pervasive prose. | 3 / 5 |
Actionability | The guidance is largely concrete and executable: specific commands (/moai plan, /clear, moai spec lint --strict), thresholds (85% coverage, 150K context), file paths (.moai/specs/SPEC-XXX/spec.md, progress.md), a copy-paste resume-message template, and fully quoted GEARS clause patterns. It stops short of 5 because several claims are unverifiable marketing ('improves task success rate by 40 percent', 'Enables 2-3x larger projects', 'Reduces escalations by 70 percent') and key steps defer to module files that are not present. | 4 / 5 |
Workflow Clarity | The three-phase workflow (plan → run → sync) is clearly sequenced with per-phase budgets, explicit /clear checkpoints, guard conditions ('Never use /clear when: In the middle of an agent task...'), a numbered multi-agent handoff procedure, and a closing verification checklist. TRUST 5 gates even specify failure actions ('Block merge and generate missing tests on failure'). Not 5 because error-recovery loops for the SPEC/DDD phases themselves (e.g. what to do when lint or coverage fails mid-run) are only pointed at via 'Detailed Reference' links rather than stated inline. | 4 / 5 |
Progressive Disclosure | Judged against the actual bundle: the body repeatedly points to nine 'modules/*.md' files (trust-5-framework.md, spec-first-ddd.md, delegation-patterns.md, token-optimization.md, progressive-disclosure.md, modular-system.md, agents-reference.md, commands-reference.md, execution-rules.md), none of which exist in the bundle — only references/examples.md and references/reference.md are present. On top of the dangling references, deep content that the skill's own rules say belongs in modules (the full GEARS spec, the closing Token Budget section duplicating modules/token-optimization.md) is inlined into SKILL.md. This matches anchor 2 ('content that clearly belongs in separate files is inlined') better than 3, since most navigation pointers lead nowhere within the bundle. | 2 / 5 |
Total | 13 / 20 Passed |