The Agent Skills Directory

[PROMPT_INJECTION]: The skill establishes an attack surface for indirect prompt injection by instructing the agent to discover external capabilities via third-party sources.
Ingestion points: The fallback strategy directs the agent to ingest content from GitHub topic pages (e.g., github.com/topics/claude-code) and general web search results.
Boundary markers: There are no instructions or delimiters provided to ensure the agent ignores or sanitizes potential malicious instructions found within discovered external skills.
Capability inventory: The skill provides templates for powerful actions, including installing Node.js packages via npx and writing new command files to the local .claude/commands/ directory.
Sanitization: The instructions do not include steps for validating the integrity or safety of external skills before recommending them to the user.
[EXTERNAL_DOWNLOADS]: The catalog references numerous integrations and MCP servers from well-known services and organizations such as Cloudflare, Vercel, Supabase, and Sentry.
[COMMAND_EXECUTION]: The skill includes shell command templates for common setup tasks, such as creating configuration directories (mkdir -p) and writing configuration data to files using echo and redirection.

external-skills-catalog