Conciseness

The five-line body is lean and assumes Claude's competence — 'Read ... and execute it exactly as written — that file is the authoritative playbook. Then follow ...' with one substitution rule — every token earns its place. No padding, no explanation of concepts Claude already knows.

Actionability

The guidance is concrete and executable in form — exact paths ('.claude/skills/loop-engineering/SKILL.md', '.agents/rules/cog.md', '.agents/agents/<name>.md') and a specific tool ('invoke_subagent'). It is not a 5 because all substantive instruction lives in an external file, so the body alone provides no examples or steps for the common cases; it is well above a 3 since nothing is pseudocode or vague.

Workflow Clarity

The sequence is clear and ordered — read and execute the authoritative playbook, then follow '.agents/rules/cog.md', with an explicit conditional rule for worker delegation. Not a 5 because there are no checkpoints clarifying how the playbook's instructions and cog.md rules interact when they conflict; not a 3 because no risky/destructive operation requires validation here.

Progressive Disclosure

References are one level deep and clearly signaled with exact paths, and splitting the substance into a shared canonical playbook is an appropriate content split for a 'shared reference' skill. It falls short of a 5 because the body is a pure pointer with no overview or summary of what the playbook contains, leaving navigation context thin.

Specificity

The description lists several specific coverage areas — 'the agent loop, deterministic verifiers, termination conditions, in-loop context management, and named patterns' — which matches the anchor for several specific actions with minor gaps. It falls short of a 5 because these are topic nouns rather than concrete actions (no verbs like extract/fill/merge), so coverage is enumerated but not action-oriented.

Completeness

Both parts are explicit: the 'what' enumerates the reference's contents ('the agent loop, deterministic verifiers, termination conditions, in-loop context management, and named patterns') and the 'when' is a concrete trigger clause ('Invoke when designing or debugging a skill that iterates'). This matches the anchor for clearly and explicitly answering both what and when with concrete trigger phrases.

Trigger Term Quality

'Invoke when designing or debugging a skill that iterates (search-verify-retry, scan-until-dry, fetch-retry-gate)' provides natural verbs users would say plus three named pattern triggers. Not a 5 because common variations such as 'retry loop', 'iteration limit', or 'verifier' as a standalone trigger phrase are absent; not a 3 because coverage goes beyond a couple of generic keywords.

Distinctiveness Conflict Risk

'loop-engineering reference for COG skills' carves out a clear niche with distinct triggers tied to iteration patterns, so overlap risk is low. It is not a 5 because generic loop/retry-related requests could plausibly match other iteration-flavored skills before this one.