Content
86%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, actionable instruction skill with concrete commands, clear sequencing, and good sectioning. The only real slack is mild redundancy across examples and unstated edge-case handling in the workflow.
Suggestions
Collapse the three Examples into one fully-worked example plus a short list of alternate triggers; the refine-confirm-save flow is already shown once.
Add an explicit 'If the user declines, do not save anything' branch to step 3/4 so the decline path is unambiguous.
Note whether to verify the learning was added (e.g., re-running caliber learn list) so the workflow has a lightweight verification step.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean with no concept-padding Claude already knows, but the three Examples repeat the same refine-confirm-save pattern; trimming to one detailed example would tighten it further. | 4 / 5 |
Actionability | Provides copy-paste-ready commands (caliber learn add, --personal, git add CALIBER_LEARNINGS.md) plus concrete type-prefix tags and worked examples covering the common cases. | 5 / 5 |
Workflow Clarity | A clear 5-step sequence with an explicit confirmation checkpoint is present, but the decline path and any post-add verification are not spelled out, leaving a minor validation gap. | 4 / 5 |
Progressive Disclosure | With no bundle files and roughly 50 lines, the skill is well organized into Instructions, Examples, and 'When NOT to trigger' sections, satisfying the simple-skill exception for one-level structure. | 5 / 5 |
Total | 18 / 20 Passed |