The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests untrusted user input describing a creator's 'Voice DNA' and interpolates it directly into multiple system prompt files (blog, twitter, linkedin, etc.). • Ingestion points: User responses to questions about their voice style, audience, and signature moves in SKILL.md. • Boundary markers: The input is placed under a '## Voice DNA' header but lacks explicit delimiters or instructions to ignore embedded commands. • Capability inventory: The skill possesses capabilities to modify local markdown and python files and execute shell commands. • Sanitization: No escaping, validation, or filtering is performed on user-provided text before interpolation.
[COMMAND_EXECUTION]: The skill uses shell commands and scripting to modify its internal logic based on user input. • Uses a python3 heredoc to rewrite channel_performance.py. • Employs perl -pi -e for global string substitution of channel IDs across skill files. • Executes yt-dlp to query YouTube channel metadata.
[CREDENTIALS_UNSAFE]: The skill prompts users for sensitive credentials, including APIFY_API_TOKEN and SUPABASE_SERVICE_ROLE_KEY, and stores them in a local .env file.
[EXTERNAL_DOWNLOADS]: Recommends the installation of external utilities yt-dlp (via pip) and youtube-transcript (via npm) to support core data fetching features.

onboard