The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted content from YouTube (transcripts and metadata) and incorporates it into the agent's context for analysis, creating a potential vector for malicious instructions to influence agent behavior.
Ingestion points: YouTube video metadata and transcripts fetched via the tools/video-analyzer.py tool as described in SKILL.md.
Capability inventory: The skill has access to the Bash and WebFetch tools and can execute local Python scripts.
Boundary markers: The instructions lack explicit boundary markers or directives for the agent to ignore potentially malicious instructions embedded within the processed transcripts.
Sanitization: No evidence of transcript sanitization or filtering is present in the skill instructions.
[COMMAND_EXECUTION]: The skill instructions mandate the automatic execution of a local script, tools/video-analyzer.py, using the Bash tool whenever a YouTube URL is encountered. The prompt explicitly directs the agent to "do not ask for permission" and "run analysis immediately," which reduces user oversight of system command execution.

pine-visualizer