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.
A well-engineered reference skill: executable API examples, a hard security gate before implementation, destructive-operation safeguards, and a clean one-level reference bundle. Its only weakness is repetition — the embed-token scoping warning and the canonical-doc consultation reminder each recur more often than needed.
Suggestions
Consolidate the token-scoping guidance (Security Checkpoint, 'Auth and scoping', Path G, 'Avoid') into one canonical statement with a single cross-reference, since it currently appears in four places.
State the 'consult createAIPrompt.md before implementing' rule once (e.g., in Doc Retrieval) instead of repeating the URL ~8 times across Paths C, D, and the canonical-sources section.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and table-driven, assumes Claude's competence (no library/API primers), and pushes detail to references ("all quality guidance lives in references/data-modeling-for-iq.md"). However, token-scoping guidance is repeated in at least four places (Security Checkpoint, "Auth and scoping", Path G, "Avoid"), and "Consult https://developer.luzmo.com/api/createAIPrompt.md" appears ~8 times — trimming opportunities that keep it below the leanest anchor. It is well above the 'some unnecessary explanation' anchor 3. | 4 / 5 |
Actionability | Guidance is copy-paste ready: a complete POST /aiprompt request with headers and JSON body, a full createAuthorization embed-token example with iq.context, per-framework component names in tables, streaming event order, region hosts, a fix table mapping each problem to exact tools (update_dataset_metadata, createFormula), and a legacy-migration table. Concrete examples cover the common cases for every path. | 5 / 5 |
Workflow Clarity | The "Choose the Right Path" routing table sequences decision-making, the [CRITICAL] Security Checkpoint is an explicit pre-implementation checklist with a stop gate ("If ANY checkbox is unchecked, STOP and fix before proceeding"), and destructive operations carry validation warnings ("Read the current AI context before overwriting it", "Confirm before update_formula or delete_formula") with user-visible change review. The Common IQ Issues table provides error→cause→fix feedback loops for recovery. | 5 / 5 |
Progressive Disclosure | All four cited bundle files exist and are one level deep, each introduced with a scope summary ("read only when relevant"); quality guidance and saved-Agent detail are deliberately split out of SKILL.md, and mcp-server.md is transparently labeled a pointer. Reference files cross-link siblings (agentic-integration ↔ saved-agents ↔ data-modeling-for-iq) at the same level, so navigation stays flat and easy. | 5 / 5 |
Total | 19 / 20 Passed |