Content
36%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.
The body reads as an auto-generated feature catalog: rich in example snippets but padded, structurally malformed (duplicate embedded frontmatter, corrupted '$' paths throughout, pseudo-syntax MCP calls), and lacking any end-to-end workflow or validation for its risky automated operations. Its broken, dangling references and total absence of a bundle mean nearly all content is inlined regardless of whether it belongs in reference files.
Suggestions
Fix the pervasive '$'-for-'/' path corruption and replace pseudo-syntax MCP blocks with actual executable tool calls or documented commands so examples are copy-paste ready.
Define a sequenced workflow (analyze repo → generate/optimize workflow → validate with a dry-run or lint step → deploy) with explicit validation checkpoints before auto-executing deploys, PRs, or issue creation.
Cut the redundant command-catalog sections (analytics, cost-optimize, failure-patterns, predict, recommend, auto-optimize) down to one representative pattern, remove the stray embedded frontmatter block, and either create real reference files for the 'See also' links or delete them.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~640-line body is heavily padded with near-identical one-command sections (analytics, cost-optimize, failure-patterns, predict, recommend, auto-optimize, profile are all variations of 'npx ruv-swarm actions X --flags'), plus a stray duplicated YAML frontmatter block with hook config that belongs in metadata. It stays above anchor 1 because it does not explain concepts Claude already knows — the bloat is redundant examples, not tutorial prose. | 2 / 5 |
Actionability | There is substantial concrete guidance (full YAML workflows, real gh CLI issue/run commands), but executability is undermined by pervasive path corruption where '$' replaces '/' (e.g. '.github$workflows$ci.yml', 'ruvnet$swarm-action@v1', 'dist$index.js') and by pseudocode MCP invocations written as if bash ('mcp__claude-flow__swarm_init { topology: "mesh", maxAgents: 12 }') or as JavaScript functions. This matches 'pseudocode instead of executable code; missing key details' rather than anchor 4, where gaps would be minor. | 3 / 5 |
Workflow Clarity | The body is a feature catalog with no overall sequence — nothing tells Claude how to actually apply this skill end-to-end — and validation is entirely absent for risky automated operations (auto-deploy with '--auto-execute', self-heal with '--auto-fix-common' creating PRs, bulk security-issue creation), which caps this dimension per the guidelines. It sits above anchor 1 only because the individual snippets are themselves coherent mini-procedures and a Best Practices section exists. | 2 / 5 |
Progressive Disclosure | Real '##'/'###' header structure provides navigation (not 'minimal structure'), but the entire ~640-line catalog is inlined monolithically with no bundle files, and the only references ('[swarm-pr.md](.$swarm-pr.md)', '[swarm-issue.md](.$swarm-issue.md)', '[sync-coordinator.md](.$sync-coordinator.md)') are buried at the very end, use corrupted '$' paths, and point to files that do not exist. This matches 'some structure; references present but not clearly signaled; content that should be separate is inline' rather than anchor 4, where references would be mostly clear and valid. | 3 / 5 |
Total | 10 / 20 Passed |