Content
78%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 dense, highly actionable protocol body with excellent token efficiency and real validation checkpoints. Its weaknesses are the dangling reference to a nonexistent KNOWLEDGE-GRAPH.md, an assumed-but-undefined section-header convention, and the absence of recovery guidance when a checkpoint fails.
Suggestions
Create the referenced KNOWLEDGE-GRAPH.md (or inline its essential entity types/triggers into SKILL.md) so the body's only external reference resolves to a real bundle file.
Define the per-agent section header convention (e.g. '## <Agent name>') that the grep query and printf append both implicitly rely on, so entries land in the correct section.
Add a brief recovery step for failed validation checkpoints — e.g. what to do when the chosen agent has a conflicting Weak entry or the familiarity path no longer exists.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is a lean ~55 lines of terse tables, format specs, and commands with zero padding or explanation of concepts Claude already knows ('Entry format: `Area | Evidence | Last Updated`', the trigger table). Every token earns its place, matching the lean anchor rather than merely 'efficient with minor trimming'. | 5 / 5 |
Actionability | Guidance is mostly executable — the grep query, printf append, and awk counter increment are real commands — but `grep -A5 "## Developer"` assumes a section-header convention never defined, and the printf append does not show how the row lands under the correct agent's section. These minor gaps keep it below fully copy-paste-ready coverage of common cases. | 4 / 5 |
Workflow Clarity | The sequence (query, delegate with context block, update on completion, append knowledge graph) is clear and backed by an explicit Validation Checkpoints section with before/after/pruning checks. However there is no error-recovery loop telling Claude what to do when a checkpoint fails, so it falls short of the feedback-loops anchor. | 4 / 5 |
Progressive Disclosure | Sections are well organized and the knowledge-graph detail is correctly split out via a clearly signaled one-level reference ('See [KNOWLEDGE-GRAPH.md](./KNOWLEDGE-GRAPH.md)'), but that referenced file does not exist anywhere in the skill bundle — the single external reference is dangling, so navigation actually breaks. Structure is present yet organization is compromised by the broken reference, fitting the middle anchor. | 3 / 5 |
Total | 16 / 20 Passed |