The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses an 'Agentic Protocol' to structure the agent's research and response process. The instructions define a persona and methodology without attempting to override system safety guidelines or extract sensitive prompt data.
[DATA_EXFILTRATION]: No sensitive file paths (e.g., credentials, SSH keys) or hardcoded secrets were found. Network tool usage is restricted to legitimate research via platform-provided tools like WebSearch and WebFetch.
[COMMAND_EXECUTION]: The provided bash scripts in the cli/ directory are interactive helpers for generating reports. They follow secure shell scripting practices, such as using printf and proper variable quoting, to prevent command injection from user-provided input.
[REMOTE_CODE_EXECUTION]: The skill does not download or execute any remote scripts or binaries.
[EXTERNAL_DOWNLOADS]: No external package installations or unauthorized downloads were detected.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user input (birth data and questions) which is processed by search tools and local scripts. The protocol includes an input classification step, and the capabilities (file writing, search) are used in a controlled, structured manner. There is no evidence of high-risk interpolation of untrusted data into sensitive contexts.

bazi-metaphysics-master