Content
35%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The skill is action-dense but far too long, padded with repetitive and partly non-executable command examples, and monolithically inline with no progressive disclosure. Destructive and batch workflows also lack validation checkpoints. It needs aggressive trimming, real executable examples, file-based reference splitting, and explicit validation steps.
Suggestions
Cut the body drastically: consolidate the repetitive `ruv-swarm github` subcommand blocks into a single command reference and remove sections that restate general knowledge (Best Practices, Metrics & Analytics, Security & Permissions).
Verify every command is executable; replace the pseudocode `mcp__claude-flow__*` JavaScript with valid, runnable invocations or remove them.
Add explicit validate->fix->retry checkpoints to the destructive/batch workflows (stale-issue closing, bulk board operations) and move large reference material into separate one-level-deep files linked from the overview.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~1240-line body is padded with repetitive near-duplicate `npx ruv-swarm github <subcommand>` blocks and prose sections (Best Practices, Metrics, Security) that largely restate what Claude already knows, fitting the verbose/padded anchor. | 1 / 3 |
Actionability | Real `gh` CLI commands are concrete and copy-paste ready, but the `mcp__claude-flow__*` JavaScript blocks are pseudocode and many `ruv-swarm github` subcommands appear non-executable, leaving guidance incomplete. | 2 / 3 |
Workflow Clarity | Sequenced workflows exist (e.g. the numbered Complete Workflow Example), but destructive/batch operations like stale-issue auto-close and bulk board ops lack validate-then-proceed feedback loops, capping this at 2. | 2 / 3 |
Progressive Disclosure | Collapsible <details> sections provide some structure, but with no bundle files the entire 1240-line reference is inline rather than split into one-level-deep reference files, which is poor for a skill this large. | 2 / 3 |
Total | 7 / 12 Passed |