Scan

Automatically discovers and installs Claude skills from SkillsMP.com, official sources, and third-party marketplaces based on project context.

When to Use

This skill is invoked when:

• /auto-skills command is run

Skill Sources

1. SkillsMP.com

Uses AI semantic search to find relevant skills.

API: GET https://skillsmp.com/api/v1/skills/ai-search

Queries based on context:

2. Official Anthropic Skills

Source: https://github.com/anthropics/skills

3. Compound Engineering (Every Marketplace)

Already installed via compound-engineering@every-marketplace

Installation Process

1. Check SKILL_INSTALL_LOCATION env var for target location
   • workspace → install to ~/.claude/skills/ (default)
   • project → install to .claude/skills/
2. Search SkillsMP using AI semantic queries
3. Review top 3 results per technology detected
4. Check for duplicates - skip if already installed
5. Install to configured location
6. Log installations for transparency

Commands Used

# Determine installation location
SKILL_DIR="${SKILL_INSTALL_LOCATION:-workspace}"
if [ "$SKILL_DIR" = "workspace" ]; then
  TARGET="$HOME/.claude/skills"
else
  TARGET=".claude/skills"
fi

# Check for existing skills in BOTH locations
get_installed_skills() {
  # Check workspace skills
  if [ -d "$HOME/.claude/skills" ]; then
    for dir in "$HOME/.claude/skills"/*; do
      if [ -d "$dir" ]; then
        basename "$dir"
      fi
    done
  fi

  # Check project skills
  if [ -d ".claude/skills" ]; then
    for dir in ".claude/skills"/*; do
      if [ -d "$dir" ]; then
        basename "$dir"
      fi
    done
  fi
}

INSTALLED_SKILLS=$(get_installed_skills)

# Search SkillsMP API
curl -X GET "https://skillsmp.com/api/v1/skills/ai-search?q=<encoded_query>" \
  -H "Authorization: Bearer $SKILLSMP_API_KEY"

# Check if skill already exists before installing
skill_name="example-skill"
if echo "$INSTALLED_SKILLS" | grep -qx "$skill_name"; then
  echo "✓ $skill_name already installed, skipping"
else
  mkdir -p "$TARGET"
  git clone <repo_url> "$TARGET/$skill_name"
fi

Example Workflow

Given project context:

{
  "languages": ["go"],
  "frameworks": ["graphql"]
}

Actions:

1. Search SkillsMP: Go development best practices performance testing
2. Search SkillsMP: GraphQL schema design resolver patterns
3. Review top 3 results from each search
4. Filter out already installed skills
5. Install up to 5 most relevant skills
6. Report what was installed

Skill Selection Criteria

When reviewing search results, prioritize skills with:

• High star count on GitHub (community approval)
• Recent updates (actively maintained)
• Clear descriptions matching the use case
• Relevant tags to detected technologies
• Good documentation (README, examples)

Safety Rules

• Max 5 new skills per auto-skills run (avoid bloat)
• Never overwrite existing skills
• Always report what was installed and why
• Skip if SKILLSMP_API_KEY is not set
• Validate skill structure before installing (must have SKILL.md)
• Check for malware in skill code before installing

Transparency

After installing, always output:

## Auto Skills Complete

Skills installed to: ~/.claude/skills/ (workspace)

Installed 3 new skills:

1. **go-performance** (⭐ 234)
   - Go performance optimization patterns
   - Source: github.com/user/go-performance

2. **graphql-schema-design** (⭐ 156)
   - GraphQL schema best practices
   - Source: github.com/user/graphql-skills

3. **testing-patterns** (⭐ 89)
   - Testing strategies for Go and GraphQL
   - Source: skillsmp.com/skills/...

Environment Variables