The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands to run local Python scripts (e.g., lyria.py, gemini_tts.py, audio_mix.py) to generate and process audio assets.
Evidence: Steps 4 and 5 utilize python3 to execute scripts located within the ${CLAUDE_PLUGIN_ROOT} directory.
Mitigation: The commands use double-quoting for arguments, but the skill should ensure that all generated text or user-provided parameters are properly escaped before being passed to the shell.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface where untrusted user input (topics, host names, tone, and music descriptions) is used to generate a script which is then processed by other AI models and CLI tools.
Ingestion points: User responses to interactive questions regarding podcast topic, host personalities, tone, and music style in SKILL.md.
Boundary markers: Absent. The skill does not explicitly instruct the agent to ignore instructions embedded within the user's provided topic or host descriptions.
Capability inventory: The skill has the capability to write and read files via audio processing scripts and interact with network-based APIs (Google Gemini/Lyria) via shell execution in SKILL.md.
Sanitization: Absent. There is no explicit logic described for sanitizing user input before it is interpolated into the script generation phase or the shell command arguments.

podcast-producer-agent