The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses local configuration files (.pawbytes/config/config.yaml) to retrieve API keys for services like fal.ai and ElevenLabs. This data is used to authenticate requests to these providers' official endpoints. While this involves sending sensitive keys to the network, it targets well-known services for the skill's primary purpose and follows standard configuration patterns.
[COMMAND_EXECUTION]: The skill extensively uses ffmpeg, curl, and the egaki CLI to process video files and interact with AI APIs. These commands are essential to the video production pipeline and are executed within the agent's expected operational scope.
[PROMPT_INJECTION]: The skill includes a research capability that ingests content from web searches to update a local knowledge base, which introduces a surface for indirect prompt injection.
Ingestion points: External data from web searches is ingested via the research-capability.md workflow into the project's knowledge directory.
Boundary markers: No explicit delimiters or boundary markers are used to isolate untrusted external content from the agent's system instructions.
Capability inventory: The skill can write files to the project directory, execute complex ffmpeg rendering commands, and perform network requests using curl and the egaki CLI.
Sanitization: There is no documentation of sanitization, filtering, or validation performed on the research findings before they are integrated into the knowledge base.

paw-cra-agent-video-producer