The Agent Skills Directory

[COMMAND_EXECUTION]: Executes shell commands via npx to interact with the skills registry (e.g., list, find, add). It also runs a specialized Python script for managing session state and journals in the user's home directory (~/.claude/discover-skills/).
[EXTERNAL_DOWNLOADS]: Performs web research across various platforms including GitHub, Reddit, and Hacker News, and queries the skills.sh registry to identify potential new skills based on identified gaps.
[REMOTE_CODE_EXECUTION]: Provides functionality to install third-party agent skills using the npx skills add command. The skill includes built-in quality verification rules, such as checking install counts and source reputation, and requires explicit user confirmation before any installation occurs.
[PROMPT_INJECTION]: Contains an indirect prompt injection surface as it ingests and processes untrusted content from the internet (web search results and registry descriptions) to provide recommendations and generate new skill proposals. * Ingestion points: Search results from Brave, GitHub, Reddit, HN, and the skills registry. * Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the teammate prompts. * Capability inventory: The skill can execute npx commands and run a local state-management script (scripts/journal-store.py). * Sanitization: No evidence of data sanitization or validation of external text before it is presented to the user or used by subagents for ideation.

discover-skills