The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a template for shell commands that directly incorporates user input. The command yt-dlp "[URL]" in SKILL.md is vulnerable to shell injection if the URL parameter is not strictly validated or escaped. An attacker could provide a crafted URL string containing command separators (e.g., ;, &, |) to execute unauthorized commands in the user's environment.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection by processing untrusted data from external sources. It transcribes audio from reels and uses the resulting text to generate a new script, creating a path for malicious instructions embedded in the audio to influence the agent.
Ingestion points: Reading transcription text from /tmp/repurpose_reel.txt (generated from an external media file via whisper).
Boundary markers: Absent. There are no specific delimiters or instructions to the LLM to ignore potentially adversarial content within the transcription.
Capability inventory: Access to shell commands (yt-dlp, whisper, rm) and local file operations within the ~/.claude/skills/repurpose/ directory.
Sanitization: No sanitization or filtering of the transcription output is performed before it is included in the LLM prompt.

repurpose