The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to search for and execute a Python script named emulo.py located in a parent directory. This involves using the shell to run python $EMULO_PY plugin profile-path --domain video, which is a high-risk pattern that bypasses standard skill isolation.- [COMMAND_EXECUTION]: The instructions include several other shell commands for file processing and verification, such as node videos/_kits/_tools/frames.mjs, ffprobe, and ffmpeg.- [DATA_EXFILTRATION]: The agent is commanded to 'Read every returned path completely' following the execution of the profile script. Since these paths are dynamically generated and supposedly contain 'personal context,' this provides a mechanism for accessing and potentially exfiltrating sensitive local files or configurations.- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process untrusted data, including a 'one line ask' brief and external reference videos.
Ingestion points: User briefs, reference films (MP4), and external film source files (JSX).
Boundary markers: None present to distinguish user-provided instructions from data.
Capability inventory: Shell command execution (python, node, ffmpeg), local file system write access, and potential network access for brand assets.
Sanitization: No evidence of input validation or escaping for the processed briefs or reference content.

video