The Agent Skills Directory

[SAFE]: The skill instructions define a legitimate workflow for project management and requirement gathering. It uses standard markdown templates to structure project information.
[DATA_EXPOSURE_AND_EXFILTRATION]: No sensitive file access or unauthorized network operations were detected. The skill reads local project files like package.json or README to gather context, which is standard and expected behavior for developer tools.
[PROMPT_INJECTION]: The instructions do not contain attempts to bypass AI safety filters, ignore system instructions, or use jailbreak-style techniques.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided answers to interview questions and writes them into files. While this creates a data ingestion surface, the instructions emphasize structured output and specific roles for the agent, providing natural boundaries against accidental instruction following from the data.
[REMOTE_CODE_EXECUTION]: No remote script downloads or dynamic code execution patterns (like eval or subprocess calls to external sources) were identified.

factory-plan