discover-skills
16
总安装量
8
周安装量
#21538
全站排名
安装命令
npx skills add https://github.com/nesnilnehc/ai-cortex --skill discover-skills
Agent 安装分布
github-copilot
7
codex
7
gemini-cli
7
opencode
7
kimi-cli
6
cursor
6
Skill 文档
Skill: Discover Skills
Purpose
Help the Agent identify missing skills for a task and recommend concrete installation steps. This skill discovers candidates and suggests what to install, but does not install or inject skills automatically.
Use Cases
- Initial bootstrap: The Agent starts with only this skill, then recommends which skills to install for the current task.
- On-demand extension: When a task requires a Skill that is not available, suggest the best matches and how to install them.
- Capability discovery: Help users find relevant skills from local indexes or public catalogs.
Behavior
- Discovery: Prefer local
skills/INDEX.mdandmanifest.jsonfor the capability map. If the user asks for external options and network access is available, search a public catalog (e.g. SkillsMP). - Matching: Match the user task to
name,description, andtags; select the best 1â3 skills. - Recommendation: Explain why each skill matches and whether it is local or external.
- Installation guidance: Provide the exact install command for each recommended skill (e.g.
npx skills add owner/repo --skill name). - Confirmation: If the user asks the Agent to install, request explicit confirmation before running any command.
Input & Output
- Input:
- Description of the current task.
- Optional: list of already installed skills.
- Optional: allowed sources (local only vs public catalogs).
- Output:
- Recommended skills with rationale.
- Install commands for each recommendation.
Restrictions
- No auto-install: Do not execute install commands without explicit user confirmation.
- No auto-injection: Do not fetch or inject remote SKILL.md content automatically.
- No bulk discovery: Avoid listing large catalogs; return only the top 1â3 matches.
Self-Check
- Relevance: Are the recommendations strongly related to the current task?
- Actionable: Are install commands concrete and correct?
- Consent: Did the Agent avoid running installs without explicit confirmation?
Examples
Example 1: Recommend a local skill
- Scenario: User asks for a standardized README.
- Steps:
- Agent checks
skills/INDEX.md. - Finds
generate-standard-readme. - Recommends it and provides:
npx skills add nesnilnehc/ai-cortex --skill generate-standard-readme.
- Agent checks
Example 2: Edge case – no local match
- Scenario: The task is niche and no local skill matches.
- Expected: Offer 1â3 external suggestions (if allowed) with install commands, or say “no match found” and ask for clarification. Do not install automatically.
Appendix: Output contract
When this skill produces recommendations, it follows this contract:
| Element | Requirement |
|---|---|
| Count | Top 1â3 matches only; no bulk catalog listing. |
| Per skill | Name, rationale (why it matches), install command (e.g. npx skills add owner/repo --skill name). |
| Source | Prefer local skills/INDEX.md and manifest.json; external only when user asks and network available. |
| Interaction | Do not run install commands without explicit user confirmation. |