The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill includes the ae_run_jsx tool, which allows the agent to execute arbitrary ExtendScript (JavaScript ES3) within the Adobe After Effects application environment. This is a powerful feature intended for operations not supported by standard tools but represents a vector for executing complex logic that could be influenced by malicious prompts.
[COMMAND_EXECUTION]: The tool ae_render_video executes the aerender command-line utility in the background. Furthermore, the skill's documentation references the use of ffmpeg for generating contact sheets from rendered frames.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from live After Effects projects, such as layer names, expressions, and markers, which may originate from untrusted sources and contain hidden instructions.
Ingestion points: Project data is read through tools like ae_comp_tree, ae_layer_detail, ae_find_animation, and ae_selection (SKILL.md).
Boundary markers: The provided instructions do not include specific delimiters or warnings to ignore potential instructions embedded within the project content.
Capability inventory: The agent has significant capabilities, including arbitrary script execution (ae_run_jsx), command-line interaction (ae_render_video), and extensive project modification rights.
Sanitization: There is no documented mechanism for sanitizing or validating project-derived data before it is used by the agent.
[CREDENTIALS_UNSAFE]: The bridge operation relies on reading an authentication token from a local configuration file at ~/.ae-mcp-bridge.json. While this is part of the bridge's functional design, accessing files containing authentication secrets is a sensitive operation.

ae-mcp-bridge