skills-discovery

AI Community

Repo stars 514
Author repo claude-plugins

Interpretation is structured for decision-making; original keeps the upstream SKILL.md unchanged.

Decide Fit First

Core job: Search for and install Agent Skills that give you specialized capabilities. Before starting work, ask might a skill exist that h…
Best fit: Use it when the task has reusable inputs, steps, and validation criteria rather than a one-off answer.
Avoid forcing it: If the source lacks commands, platform support, or external-service evidence, keep those fields unknown instead of guessing.

Design Intent

Structure: The skill is organized around “When to search for skills”, “Discovery workflow”, “Search strategies”, “Installation”, showing how the author expects the agent to judge fit, collect context, and produce verifiable output.
Trigger evidence: Prioritize the author’s wording around when to use it, what context to collect, and what output shape to produce.
Evidence boundary: Author text states facts, repository files prove commands and paths, and Fluxly only adds fit, limits, and usage judgment.

How To Use It

Inputs: Provide target material, scope, expected result, forbidden changes, and validation method.
Invocation: Name skills-discovery directly; if the source includes slash commands, start with the command and then add task context.
Validation: Start small and check whether the result follows “When to search for skills / Discovery workflow / Search strategies” before expanding.

Boundaries And Review

Dependencies: It usually needs no extra API key, so start with a small validation task.
Permissions: Declared permissions include read / write; ask the agent to state file, command, and rollback boundaries before acting.
Quality bar: A useful result names the deliverable, evidence, and next action. Generic prose means the task needs tighter context.

Skills Discovery

You can extend your capabilities by discovering and installing Agent Skills from the claude-plugins.dev registry. Skills provide specialized knowledge, tools, and techniques for specific technologies, frameworks, and domains.

When to search for skills

First, check if an installed skill matches the task. If not, search the registry—specialized skills may exist that you haven't installed yet.

Before starting any non-trivial task, ask yourself:

Do I have a skill for this? → Use it
Might one exist that I don't have? → Search the registry

Search proactively when:

The task involves specific technologies, frameworks, or file formats
You're about to do something where best practices matter (testing, deployment, APIs, documentation)
The domain is specialized (PDF processing, data pipelines, ML workflows)
You notice yourself about to give generic advice where expert patterns would help

Also search when users explicitly ask to find, install, or manage skills.

Discovery workflow

Use the registry API for search (the CLI's search command is interactive and not suitable for programmatic use):

curl "https://claude-plugins.dev/api/skills?q=QUERY&limit=20&offset=0"

Parameters:

q: Search query (e.g., "frontend", "python", "pdf")
limit: Results per page (max 100)
offset: Pagination offset

Response structure:

{
  "skills": [
    {
      "id": "...",
      "name": "skill-name",
      "namespace": "@owner/repo/skill-name",
      "sourceUrl": "https://github.com/...",
      "description": "...",
      "author": "...",
      "installs": 123,
      "stars": 45
    }
  ],
  "total": 100,
  "limit": 10,
  "offset": 0
}

Search strategies

The registry indexes skill names, descriptions, and tags. Construct queries that match how skill authors describe their work.

Query construction:

Use 1-3 specific terms (too broad = noise, too narrow = misses)
Prefer widely-used terminology over project-specific jargon
Technology + task often outperforms either alone
If results are poor, broaden or try synonyms

Installation

Determine which client the user is working in before installing. If unclear, ask.

Supported clients:

claude-code — Claude Code CLI
codex — Codex
cursor — Cursor editor
amp - amp CLI
opencode - OpenCode CLI
goose - Goose CLI
github — VSCode/ github
vscode — VS Code
letta — Letta CLI
gemini - Gemini CLI
windsurf - Windsurf editor
antigravity - Antigravity
trae - Trae
qoder - Qoder
codebuddy - CodeBuddy
Client selection:

npx skills-installer install @owner/repo/skill-name --client claude-code  # default
npx skills-installer install @owner/repo/skill-name --client cursor
npx skills-installer install @owner/repo/skill-name --client vscode

Scope selection:

npx skills-installer install @owner/repo/skill-name  # global (default)
npx skills-installer install @owner/repo/skill-name --local  # project-specific

Combined:

npx skills-installer install @owner/repo/skill-name --client cursor --local

Defaults:

Client: claude-code
Scope: global

Management

# List installed skills
npx skills-installer list

# Uninstall a skill
npx skills-installer uninstall @owner/repo/skill-name

Presenting results to users

When you find relevant skills:

Show 3-5 most relevant results maximum
Include: name, namespace, description, stars, installs
Explain how each skill helps with their specific task
Prioritize those with high installs
Always ask for confirmation before installing
Offer to help directly if no good skill exists or user declines

Examples

Example: Proactive suggestion

User: "I need to create a Django REST API"

curl "https://claude-plugins.dev/api/skills?q=django&limit=10"

Present suggestion:

I found some skills that could help:

1. django-rest-framework-expert (@anthropics/claude-code/django-rest-framework-expert)
   Description: Django REST API development with best practices
   ⭐ 234 stars • 1,567 installs

Would you like me to install this, or help you directly without installing a skill?

Example: Explicit search request

User: "find skills for Python"

curl "https://claude-plugins.dev/api/skills?q=python&limit=10"

Present results and ask which to install.

API reference

Endpoint	Description
`GET /api/skills/search?q=QUERY`	Search skills
`GET /api/skills/@owner/repo/skill-name`	Get skill details

Web registry: https://claude-plugins.dev/skills

Troubleshooting

No results found:

Try broader search terms
Browse web registry: https://claude-plugins.dev/skills

Installation fails:

Verify namespace format: @owner/repo/skill-name
Check skill exists in registry
Verify directory permissions

Skill not activating:

User may need to restart their client
Verify correct installation directory
Confirm SKILL.md exists in installation path

Fluxly profile Author and license come from source; runtime, permissions, and network are Fluxly detections or estimates

Fluxly category: AI
Author-declared agents: No explicit declaration found; this is not inferred or tested compatibility
Static check: 88 / 100 · heuristic scan, not runtime safety proof
Author / version / license: @Kamalnrf · no license declared
Fluxly token estimate: Lean
Fluxly setup estimate: Plug-and-play
External API key: No requirement detected
Detected OS requirements: macOS · Linux · Windows
Runtime requirements: Python
Detected file/system behavior: Read-only
Write / modify
Detected network behavior: External requests
Install commands: None (reference only)

Profile is derived at build time from SKILL.md and install vectors. Subject to drift from author intent.

Heads up: 未限定 allowed-tools，默认拥有全部工具权限。

Output preview skills-discovery.preview

# Presenting results to users

1. Show 3-5 most relevant results maximum
2. Include: name, namespace, description, stars, installs
3. Explain how each skill helps with their _specific_ task
4. Prioritize those with high installs
5. Always ask for confirmation before installing
6. Offer to help directly if no good skill exists or user declines

SKILL.md SKILL.md

# Skills Discovery

You can extend your capabilities by discovering and installing Agent Skills from the claude-plugins.dev registry. Skills provide specialized knowledge, tools, and techniques for specific technologies, frameworks, and domains.

## When to search for skills

First, check if an installed skill matches the task. If not, search the registry—specialized skills may exist that you haven't installed yet.

Before starting any non-trivial task, ask yourself:

1. Do I have a skill for this? → Use it
2. Might one exist that I don't have? → Search the registry

Search proactively when:

- The task involves specific technologies, frameworks, or file formats
- You're about to do something where best practices matter (testing, deployment, APIs, documentation)
- The domain is specialized (PDF processing, data pipelines, ML workflows)
- You notice yourself about to give generic advice where expert patterns would help

Also search when users explicitly ask to find, install, or manage skills.

## Discovery workflow

Use the registry API for search (the CLI's search command is interactive and not suitable for programmatic use):

```bash
curl "https://claude-plugins.dev/api/skills?q=QUERY&limit=20&offset=0"
```

**Parameters:**

- `q`: Search query (e.g., "frontend", "python", "pdf")
- `limit`: Results per page (max 100)
- `offset`: Pagination offset

**Response structure:**

```json
{
  "skills": [
    {
      "id": "...",
      "name": "skill-name",
      "namespace": "@owner/repo/skill-name",
      "sourceUrl": "https://github.com/...",
      "description": "...",
      "author": "...",
      "installs": 123,
      "stars": 45
    }
  ],
  "total": 100,
  "limit": 10,
  "offset": 0
}
```

## Search strategies

…

Evidence boundary and execution chain

Author text anchors workflow facts; Fluxly only indexes current sections, terms, files, and commands.

sections -> When to search for skills → Discovery workflow → Search strategies → Installation → Management → Presenting results to users
terms    -> Parameters · Response structure · Query construction · Supported clients · Client selection · Scope selection · Combined · Defaults
files/cmd -> limit · offset · claude-code · amp · opencode · goose · vscode · letta
body sha256 -> c6f938a06102