The Agent Skills Directory

[COMMAND_EXECUTION]: The Python and PowerShell scripts execute local system commands to perform their primary management functions.
scripts/search_skills.py and scripts/Search-Skills.ps1 invoke the gh (GitHub CLI) tool to perform code and repository searches, fetch metadata via API, and list file contents.
The scripts use curl to download skill-related files from remote repositories to the local user directory.
[EXTERNAL_DOWNLOADS]: The skill downloads content from external sources to manage its skill index and facilitate installations.
It fetches SKILL.md files, scripts, and documentation from external GitHub repositories.
These operations target established domains including github.com and raw.githubusercontent.com.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, which presents a surface for indirect prompt injection.
Ingestion Points: The tool fetches repository names, descriptions, and markdown content from arbitrary GitHub repositories found through the search function.
Boundary Markers: The agent is instructed in SKILL.md and references/agent-instructions.md to show source URLs and obtain explicit user approval before installation, which helps mitigate the risk of accidental execution of malicious content.
Capability Inventory: The skill scripts have the capability to write to the local file system (for installation) and perform network requests (via gh and curl).
Sanitization: Web search queries are URL-encoded before being displayed to the user to prevent basic injection into the generated search URLs.

skill-finder