sensei
**WORKFLOW SKILL** — Iteratively improve skill frontmatter compliance using the Ralph loop pattern. WHEN: "run sensei", "sensei help", "improve skill", "fix frontmatter", "skill compliance", "frontmatter audit", "score skill", "check skill tokens". INVOKES: token counting tools, test runners, git commands. FOR SINGLE OPERATIONS: use token CLI directly for counts/checks.
81
76%
Does it follow best practices?
Impact
Pending
No eval scenarios have been run
Passed
No known issues
Optimize this skill with Tessl
npx tessl skill review --optimize ./.github/skills/sensei/SKILL.mdQuality
Discovery
89%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This is a well-structured skill description with excellent trigger term coverage and clear completeness. Its main weakness is moderate specificity — while it names the domain and general goal, the concrete actions performed during the iterative improvement loop could be more explicitly listed (e.g., 'validates fields, checks token limits, suggests fixes'). The 'Ralph loop pattern' reference is internal jargon that doesn't add clarity for selection purposes.
Suggestions
Replace or supplement 'Ralph loop pattern' with a brief enumeration of concrete steps (e.g., 'validates required fields, checks token limits, scores compliance, suggests fixes').
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Names the domain (skill frontmatter compliance) and some actions ('improve', 'audit', 'score', 'check tokens'), but the core action 'iteratively improve skill frontmatter compliance using the Ralph loop pattern' is somewhat abstract — 'Ralph loop pattern' is internal jargon and the specific concrete steps aren't enumerated. | 2 / 3 |
Completeness | Clearly answers both 'what' (iteratively improve skill frontmatter compliance) and 'when' (explicit WHEN clause with trigger phrases). Also includes an INVOKES section and a disambiguation note for single operations, which adds useful context for skill selection. | 3 / 3 |
Trigger Term Quality | Includes a strong set of natural trigger terms that users would actually say: 'run sensei', 'sensei help', 'improve skill', 'fix frontmatter', 'skill compliance', 'frontmatter audit', 'score skill', 'check skill tokens'. These cover multiple natural phrasings and variations. | 3 / 3 |
Distinctiveness Conflict Risk | Highly distinctive — the 'sensei' naming, 'Ralph loop pattern', and focus on skill frontmatter compliance create a very clear niche. The disambiguation note ('FOR SINGLE OPERATIONS: use token CLI directly') further reduces conflict risk with related tools. | 3 / 3 |
Total | 11 / 12 Passed |
Implementation
62%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The skill provides excellent actionability and workflow clarity with a well-defined iterative loop, concrete commands, and clear validation checkpoints. However, it suffers significantly from verbosity — the same concepts (especially 'DO NOT USE FOR' risks and the loop steps) are repeated multiple times, and content that should live in reference files is duplicated inline. The ASCII help box alone consumes substantial tokens for information that's restated in the body.
Suggestions
Remove the large ASCII help box or move it to a reference file — the invocation modes and workflow steps already cover this information more concisely in the body.
Consolidate the three near-identical 'DO NOT USE FOR' warnings into a single concise statement, referencing SCORING.md for the full rationale.
Move the detailed GEPA explanation (step 5b) to a reference file like references/GEPA.md and keep only a one-line summary with a link in the main skill.
Remove the inline scoring criteria table and quick reference since SCORING.md already exists as a reference — just link to it once.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is extremely verbose at ~200+ lines. The large ASCII art help box, repeated explanations of 'DO NOT USE FOR' risks (stated 3 times with nearly identical content), duplicate loop descriptions (once in the help box, once in the main section), and extensive caveats all waste tokens. Much of this content could be in reference files rather than inline. | 1 / 3 |
Actionability | The skill provides concrete, executable commands throughout: specific CLI invocations, bash commands for scaffolding, exact test runner commands, GEPA script paths, and commit message formats. The Ralph loop steps include specific commands at each stage. | 3 / 3 |
Workflow Clarity | The 12-step Ralph loop is clearly sequenced with explicit validation checkpoints (step 3 checks score before proceeding, step 7 runs tests, step 8 validates references, step 12 caps iterations at 5). There's a clear feedback loop: iterate until score >= Medium-High AND tests pass, with max 5 iterations. | 3 / 3 |
Progressive Disclosure | Reference files are well-signaled (SCORING.md, LOOP.md, EXAMPLES.md, TOKEN-INTEGRATION.md), but too much content that belongs in those reference files is duplicated inline — the scoring criteria, the full loop details, and the GEPA explanation are all expanded in the main file despite having dedicated reference documents. | 2 / 3 |
Total | 9 / 12 Passed |
Validation
100%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 11 / 11 Passed
Validation for skill structure
No warnings or errors.
- Repository
- microsoft/github-copilot-for-azure
- Commit
a46a937
- Reviewed
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.