The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted content from script and transcript files, creating a surface for indirect prompt injection. 1. Ingestion points: The agent reads script.md and takes_packed.md from the workspace folder. 2. Boundary markers: No explicit delimiters or instructions are provided to the agent to treat this data as untrusted or to ignore embedded commands. 3. Capability inventory: The skill is configured with Read, Write, and Bash tools. 4. Sanitization: No input sanitization or validation of the text content is performed.
[COMMAND_EXECUTION]: The skill invokes a local Python utility through the shell. 1. Evidence: The process instructions require calling helpers/timeline_view.py with parameters derived from the processed video data. 2. Context: This execution is limited to local scripts and is used specifically for resolving ambiguities in edit points.

edit-director