Content
88%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 strong, tightly engineered instruction skill: the 8-step workflow is sequenced with real validation checkpoints and concrete thresholds, and the description-adjacent scope guards prevent automation and metric-gaming failures. The main costs are repetition of the compliance and registry rules across sections, and a reference graph that points entirely outside the shipped bundle, which weakens standalone navigability.
Suggestions
State the zero-automation red line and the registry propose rule once (e.g., in the Scope guard and Skill Contract) and reference them from other sections instead of repeating near-verbatim in Instructions, Save Results, and Done-when.
Ship or inline the load-bearing referenced material — the handoff format from skill-contract.md and the propose-request format for registry-events.py are needed to execute steps 7-8 but live in files outside the bundle.
Add a brief note (or restructure paths) so the skill is self-contained when cloned outside the parent repo, since all ten reference links currently depend on the ../../.. directory layout.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is efficient and almost entirely domain-specific knowledge Claude cannot infer (registry paths, Measured/User-provided/Estimated labeling, give:ask ledger, ECHO discipline), with no explanation of concepts Claude already knows. It is not 5 because the same facts are restated multiple times — the zero-automation red line appears in the scope guard, the Done-when list, and step framing, and the registry-propose rule is repeated nearly verbatim in Skill Contract, step 7, and Save Results — which could be trimmed. | 4 / 5 |
Actionability | For an instruction-only skill the guidance is fully executable: concrete numbers (30-45 min block, "3+ substantive touches across 2+ weeks", "~1 week" funding-congratulations decay), exact paths (memory/social/social-selling-planner/), a named command path (registry-events.py with operation: propose), a NEEDS_INPUT stop condition naming exactly what to provide, and copy-paste Quick Start prompts. It is not 4 because the common cases (build block, fired trigger, quarterly diagnostic) are each covered with specific, executable parameters rather than minor gaps. | 5 / 5 |
Workflow Clarity | Eight clearly sequenced steps with explicit validation checkpoints: verify each trigger "at its source and label it Measured with URL", a "Done when" checklist of five criteria, a NEEDS_INPUT stop when no target list exists, and termination rules with a visited-set check and ambiguity stop. It is not 4 because checkpoints are explicit rather than implicit, and error/ambiguity paths (stop conditions, ask-before-write, flagging sub-threshold asks as cold pitches) are stated rather than missing. | 5 / 5 |
Progressive Disclosure | The SKILL.md is a well-structured overview with an annotated Reference Materials section and one-level-deep links, keeping detail out of the main file. It is not 5 because no bundle files ship with the skill — there are no references/, scripts/, or assets/ directories — and every reference resolves outside the skill directory (../../../references/echo-benchmark.md, scripts/connectors/bluesky.py, ../../observe/social-pulse-monitor/SKILL.md), so navigation depends on an unstated surrounding repo layout and none of the referenced paths can be verified from the bundle. | 4 / 5 |
Total | 18 / 20 Passed |