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 tight, self-contained satellite skill: an unambiguous decision table with tie-break rules, concrete exemplars for every mode, and a complete spawn template. The only weaknesses are minor — mixed placeholder conventions and an unexplained Jinja conditional in the template, and a large inline template that could live in a bundle file.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence: a dense selection table, one-line exemplars ("Fix the typo in README → Spawn one agent, no charter, no history read"), compact upgrade rules, and a template with no filler or explanation of concepts Claude already knows. Only trivially trimmable fragments ("A user favorite — make it instant") exist, which does not drop it to the 4 anchor's "minor instances of over-explanation". | 5 / 5 |
Actionability | Highly actionable: the decision table gives concrete criteria and targets, each mode has copy-ready exemplar prompts, and the Lightweight Spawn Template is a complete fill-in prompt including the read-only variant ("agent_type: 'explore' with 'You are {Name}...'"), covering the common cases. Not a 5 because of minor gaps: the Jinja "{% if WORKTREE_MODE %}" block is used without explanation, and placeholder conventions are inconsistent ("{resolved_model}" vs "<resolved CURRENT_DATETIME literal>") with no resolution guidance. | 4 / 5 |
Workflow Clarity | The procedure is fully determined and unambiguous: "After routing determines WHO handles work, select the response MODE based on task complexity", the table fixes base selection, the upgrade rules act as explicit tie-break checkpoints ("If uncertain between Lightweight and Standard → choose Standard"), and "Never downgrade mid-task" closes the loop. No destructive or batch operations are involved, so no validation cap applies; the decision checkpoints match the 5 anchor. | 5 / 5 |
Progressive Disclosure | Good structure with clear sections (selection table, per-mode exemplars, upgrade rules, spawn template) and no nested references — the only external mention, the "1-line stub in squad.agent.md", is clearly signaled as awareness-only. Not a 5 because the ~28-line spawn template is inlined where a references/ file could carry it, and the body exceeds the under-50-lines simple-skill case; not a 3 because nothing that clearly belongs elsewhere is buried and navigation is easy. | 4 / 5 |
Total | 18 / 20 Passed |