网有NETHAVE
AGENT SKILL

sensei

**WORKFLOW SKILL** - Improve and evaluate Agent Skills with portable validation, host profiles, resource analysis, and behavioral evidence. WHEN: "run sensei", "sensei help", "improve skill", "fix frontmatter", "score skill", "analyze skill", "evaluate skill". INVOKES: Sensei CLI, test runners, git. FOR SINGLE OPERATIONS: use the Sensei CLI directly.

使用场景测试与质量Skill与智能体开发
License 许可
MIT
所属仓库
spboyer/sensei
最近更新
2026-11-14
Stars
★ 60
Forks
8

01 前言 Front Matter

name
sensei
description
**WORKFLOW SKILL** - Improve and evaluate Agent Skills with portable validation, host profiles, resource analysis, and behavioral evidence. WHEN: "run sensei", "sensei help", "improve skill", "fix frontmatter", "score skill", "analyze skill", "evaluate skill". INVOKES: Sensei CLI, test runners, git. FOR SINGLE OPERATIONS: use the Sensei CLI directly.

02 资源文件 158

03 SKILL.md 正文

Sensei

Improve skills using the Ralph loop. For help, explain modes, defaults, and checks from configuration. For source bundles and installation, see distribution.

  1. Read the target SKILL.md, references, tests, and repo-local instructions. Verify tooling exists before claiming it is missing. Count baseline tokens: From the Sensei checkout, pass the target repository root and a path relative to it: npm run tokens -- count --root <target-repo> <skill-path>/SKILL.md.
  2. Score with npm run tokens -- score --root <target-repo> <skill-path>. Select --profile=portable or --profile=copilot explicitly when host extensions are present. Separate portable-spec diagnostics from Sensei recommendations. Preserve .sensei.json and AGENTS.md short-description policy. A 150-character floor, 60-word target, type prefix, WHEN:, and 500-token target are heuristics, not specification requirements.
  3. Scaffold missing tests using templates. Use realistic positive prompts and close near-misses from intended scope; negative frontmatter syntax is not required.
  4. Improve concise positive routing and operational instructions using examples. Keep descriptions non-empty and within 1024 characters. Preserve meaningful boundaries; evaluate them on the target host. Check portability before introducing host fields.
  5. Verify with the target project's test runner; --fast skips tests but must report them as not run. GEPA mode uses optional keyword/proxy fitness, not observed host activation; follow configuration.
  6. Analyze resources with npm run tokens -- analyze --root <target-repo> <skill-path>. Review contained links, symlinks, aggregate resource limits, script signals, and lifecycle evidence. Treat heuristic warnings as hypotheses, not failures.
  7. Check budgets with npm run tokens -- check --root <target-repo> <skill-path>/SKILL.md. Report configured token-limit warnings separately from optional optimization suggestions. For MCP-aware skills, follow MCP integration.
  8. Summarize profile states, findings, unavailable evidence, changes, diagnostics, tests, token deltas, and remaining suggestions. score --emit-proof records static evidence, not runtime quality or a security audit. Ask whether to commit, create an issue, or skip; do not discard edits unasked.
  9. Repeat until the intended gates and tests pass, at most five iterations. Process batches sequentially and report failures explicitly.

04 同仓库更多 Skills

spboyer
spboyer/sensei

sample-skill

Utility for working with files

Skill与智能体开发
★ 60
spboyer
spboyer/sensei

squad-conventions

Core conventions and patterns used in the Squad codebase

★ 60
spboyer
spboyer/sensei

{skill-name}

{what this skill teaches agents}

测试与质量Skill与智能体开发
★ 60