Content
92%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.
An exemplarily structured skill body: concrete commands that match the shipped script, a clearly sequenced workflow with an explicit validation gate and verification checklist, and one-level-deep references with per-file load triggers. The only notable improvement is mild redundancy between the Mindset principles and the Anti-Patterns' WHY clauses.
Suggestions
Trim or merge the Mindset section into Core Rules/Anti-Patterns: its unevidenced-status and asymmetric-trust principles are restated almost verbatim as Anti-Patterns' WHY clauses (e.g., 'NEVER close an outside-reach item on your own observation').
In the description, scope the generic completion phrases ("are we done", "close a goal out") to the session-goal context, e.g. "are we done with the session goal", to reduce overlap with plan/todo-tracking skills.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and assumes competence — no concept explanations, just rules, commands, and pointers — but the Mindset section ("An unevidenced status is worse than no status... Keep trust asymmetric") partially restates the rationale already carried by the Anti-Patterns' WHY clauses, so a small amount could be trimmed. | 4 / 5 |
Actionability | Fully executable, copy-paste-ready commands with argument shapes cover the common cases: "./scripts/goal.sh new \"Rotate the exposed token and unblock Phase 1\"", "./scripts/goal.sh park 3 \"Andy is on leave until Monday\"", "./scripts/goal.sh check --session <id>", plus a verbatim status output example and a concrete BAD/GOOD row comparison — all matching the actual goal.sh interface. | 5 / 5 |
Workflow Clarity | The sequence is explicit (clarity test → draft → user confirmation → in-progress with same-turn evidence → park → status → promote → close) with an explicit validation gate ("Close out only when the check exits zero"), a five-item verification checklist, and a signaled recovery path (references/file-shape.md "load when diagnosing a failing check"). | 5 / 5 |
Progressive Disclosure | SKILL.md is a concise overview with five one-level-deep bundle references, each existing on disk and carrying an explicit load condition ("load when a new goal lacks a checkable end state", "load when closing any item", etc.), plus a schema asset and the goal.sh script — well-signaled, easy navigation. | 5 / 5 |
Total | 19 / 20 Passed |