The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes the ffmpeg binary and internal Python scripts via subprocess.run to perform video encoding and image rendering. These executions are intended for the skill's media generation functionality.
[DYNAMIC_EXECUTION]: The tool uses the importlib library to dynamically load its own parsing logic (scripts/extract.py) during execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes conversation logs which act as a source of untrusted data. A malicious conversation history could potentially contain instructions aimed at the agent processing the files.
Ingestion points: Conversational data is read from local ChatGPT exports, ~/.claude/projects, and ~/.codex/sessions.
Boundary markers: The skill does not employ delimiters or specific instructions to the agent to ignore content within the processed chat logs.
Capability inventory: The skill has access to the local filesystem and can execute system commands through the ffmpeg tool.
Sanitization: The extraction logic performs basic character truncation on messages but lacks comprehensive filtering of prompt injection patterns.

ai-dna