The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted content from YouTube transcripts.
Ingestion points: Video transcript text is extracted from the DOM via scripts/extract-transcript-segments.py.
Boundary markers: The instructions in SKILL.md do not define clear delimiters or use explicit guardrails to prevent the agent from following instructions embedded within the transcript text.
Capability inventory: The skill utilizes a bash tool to execute local Python scripts and reads/writes to a local memory file in {working-directory}/browser-act-skill-forge-memories/.
Sanitization: There is no evidence of filtering or sanitization of the extracted transcript text before it is processed by the agent.
[COMMAND_EXECUTION]: The skill uses eval within the documentation to describe how to run local scripts (e.g., eval "$(python scripts/get-languages.py)"). While eval can be dangerous, here it is used to execute trusted, static local files provided with the skill to generate browser automation code.

youtube-transcript