The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted input from external Slack, GitHub, and Email sources, presenting a surface for indirect prompt injection (Category 8). * Ingestion points: Source connectors for Slack, GitHub, and Email as described in SKILL.md and persona files. * Boundary markers: The skill contains clear instructions in SKILL.md and source references to treat external content as data and not instructions. * Capability inventory: Access to Bash, git, gh, curl, and media processing tools. * Sanitization: The agent is instructed to summarize feedback rather than quote it verbatim, and scripts include validation for fix references.- [COMMAND_EXECUTION]: The skill executes local scripts and system utilities to manage state and process media attachments. * Evidence: Setup commands in SKILL.md and subprocess calls to ffmpeg and ffprobe in analyze_riffrec_zip.py.- [EXTERNAL_DOWNLOADS]: The skill interacts with the OpenAI API for media transcription. * Evidence: The script analyze_riffrec_zip.py uses curl to communicate with api.openai.com. This is a well-known service and the skill correctly manages authentication via environment variables.

ce-sweep