Content
100%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.
An exemplary command-reference skill: lean, fully executable, with a well-sequenced long-running-task pattern including an explicit status checkpoint and a useful 'When NOT to use' boundary section. Nothing needs to be split out or trimmed.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The two-sentence intro conveys only non-obvious environment facts ('on PATH in the gateway', 'persist across tool-call boundaries'), and every command carries a single-purpose comment with zero padding — every token earns its place. | 5 / 5 |
Actionability | All commands are copy-paste executable and cover the common cases (new-session -d, capture-pane with -S -1000 scrollback, send-keys with the C-m note), and the status-flag pattern is a complete, correctly escaped executable script. | 5 / 5 |
Workflow Clarity | Each single action is unambiguous (simple-skill case), and the flagship pattern sequences start → later check of /tmp/build.exit → report status or tail the log, providing an explicit checkpoint with no validation gaps. | 5 / 5 |
Progressive Disclosure | No bundle files exist and none are needed at this scope — all content is core overview material appropriately inlined, organized into clear one-level sections with easy navigation. | 5 / 5 |
Total | 20 / 20 Passed |