The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest local project data, such as README.md, guides, and source code (including authentication and onboarding models), to generate synthetic user personas. This represents an indirect prompt injection surface where malicious instructions embedded in project files could theoretically influence persona generation. The skill mitigates this risk through evidence-based synthesis rules, confidence thresholds, and a requirement for human-in-the-loop research validation before promoting personas to an 'active' lifecycle status.
[COMMAND_EXECUTION]: The documentation for the SPEAK mode includes shell commands for the agent to verify the availability of text-to-speech engines. These include curl requests to a local VOICEVOX server (http://localhost:50021/version), which to locate binaries, and a Python one-liner to check for the Google Cloud Text-to-Speech library. These commands are diagnostic in nature and intended for local environment verification.
[EXTERNAL_DOWNLOADS]: The environment check for the SPEAK mode utilizes npx --yes edge-tts, which can trigger the automatic download and execution of the edge-tts package from the NPM registry if it is not already present on the system. This is a standard dependency management pattern for Node.js utilities and targets a well-known package.

cast