Content
85%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 unusually well-built router skill: concrete routing (specific files, error codes, CLI commands), unambiguous decision logic with explicit conflict resolution against sibling suites, and ideal progressive disclosure that delegates all detail to one-level-deep sub-skill files. Its only real weakness is redundancy — the routing table, dot-graph tree, numbered tree, and example invocations restate the same mappings, inflating token cost without adding guidance.
Suggestions
Delete one of the two Decision Tree renderings — the dot graph and the numbered list encode identical routing; keep the numbered list (cheaper to scan) and cut ~20 lines.
Trim Example Invocations to the handful of phrasings not already covered by Quick Reference rows verbatim (e.g. keep the lock-screen intents and scanner cases, drop 'How do I store an auth token securely?' which restates an existing table row).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every line is information-dense with no concept explanations Claude already knows, but the same symptom-to-file routing is rendered up to four times: the Quick Reference table, the dot-graph Decision Tree, the numbered Decision Tree that fully duplicates the graph, and 13 Example Invocations that restate table rows (e.g. 'How do I store an auth token securely?' vs. the 'Store tokens, passwords, API keys securely' row). Roughly 60 of ~245 lines could be cut, which is 'mostly efficient but could be tightened' rather than the 'minor instances' of anchor 4 or the padded explanations of anchor 2. | 3 / 5 |
Actionability | Routing guidance is fully concrete: exact paths ('skills/keychain-diag.md'), specific error-code triggers ('errSecDuplicateItem', 'errSecItemNotFound', 'errSecInteractionNotAllowed', 'CODESIGN', 'ITMS-90xxx'), named CLI commands ('security find-identity', 'security cms -D', 'codesign -d --entitlements'), and explicit agent-launch instructions (axiom_get_agent over MCP, or inline fallback via file search). Per the rubric's instruction-only note, absence of code is not penalized when guidance is actionable, and every common case has an unambiguous next action. | 5 / 5 |
Workflow Clarity | This is a single-action router skill — diagnose the task, route to one file — and that action is unambiguous: the Quick Reference, Decision Tree, and Conflict Resolution rules ('CODESIGN, ITMS-90xxx, errSec → security') triangulate every case including boundary disputes with sibling suites. Per the simple-skill guidance a single unambiguous action earns the top score, and the skill involves no destructive or batch operations that would require validation checkpoints. | 5 / 5 |
Progressive Disclosure | Exemplary router structure: no reference detail is inlined, and everything is delegated to clearly signaled one-level-deep files ('skills/keychain.md', 'skills/cryptokit-ref.md', 'skills/code-signing-diag.md') with a symptom-to-file Quick Reference for navigation and cross-suite pointers labeled with both suite and path ('See axiom-data (skills/grdb-app-groups.md) §4'). The referenced files live in the parent suite and were not present in this review bundle to verify, but the structure as written matches the anchor-5 pattern: concise overview plus well-signaled, one-level-deep references. | 5 / 5 |
Total | 18 / 20 Passed |