Content
81%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-structured, evidence-disciplined instruction skill: clear contract, gated data requirements, an 8-step template-driven workflow with explicit completion criteria and a NEEDS_INPUT feedback loop, and genuine keyless connector recipes. Its main flaws are minor: duplicated next-skill navigation across three sections and heavy reliance on out-of-bundle files for the finer mechanics.
Suggestions
Consolidate the navigation sections: 'Primary next skill' (Skill Contract), 'Reference Materials' sibling list, and 'Next Best Skill' all cover the same links — keep one 'Next Best Skill' section and drop the duplicates to recover tokens.
Inline or excerpt the exact 'registry-events.py' operation: propose request shape (even a one-line example payload) so the promotion step is as executable as the connector calls.
Clarify the status of out-of-bundle references (skill-contract.md, state-model.md, CONNECTORS.md, sibling skills) — either confirm the paths resolve in the containing repo or inline the one or two rules (e.g., the Handoff Summary shape) the skill depends on.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and assumes competence (contract semantics, authorization rules, connector recipes are stated once without tutorial padding), but navigation is duplicated: 'Primary next skill: campaign-planner' in the Skill Contract is repeated in 'Next Best Skill', and the Scout siblings are listed in both 'Reference Materials' and 'Next Best Skill'. Not 5: that redundancy is trimming that would earn back tokens; not 3: the padding is minor, not 'several unnecessary explanations'. | 4 / 5 |
Actionability | Concrete, executable commands are present — 'python3 "${CLAUDE_PLUGIN_ROOT}/scripts/connectors/gdelt.py" '"<competitor>"' --days 30' and the keyless 'https://www.youtube.com/feeds/videos.xml?channel_id=UC…' RSS recipe piped into rss_monitor.py — plus per-step template links and an exact save path. Not 5: the referenced connector scripts and registry-events.py workflow sit outside this bundle and their invocation is only sketched, leaving a gap for the most complex operations; not 3: what is given is directly executable, not pseudocode. | 4 / 5 |
Workflow Clarity | Eight numbered steps, each anchored to a fill-in template; a three-item 'Done when' checklist; and an explicit feedback loop — missing evidence returns 'NEEDS_INPUT' with a collection plan, and estimates stay 'TBD/NEEDS_INPUT' until the formula and inputs are declared, which the Worked Example demonstrates. Not 4: validation checkpoints are explicit and complete rather than mostly present. | 5 / 5 |
Progressive Disclosure | The single in-bundle reference (references/templates.md) is one level deep, well-signaled per step, and its anchors were verified to exist; the Reference Materials section annotates every pointer. Not 5: several references point outside the skill directory ('../../../references/skill-contract.md', sibling skills) where their content cannot be confirmed, and the next-skill navigation is stated in two places — minor organization gaps. | 4 / 5 |
Total | 17 / 20 Passed |