The Agent Skills Directory

[COMMAND_EXECUTION]: Employs standard shell utilities including git, date, find, ls, and mkdir to navigate the project structure and retrieve metadata for analysis.
[SAFE]: Operates within a designated local directory structure (ψ/memory/) for persistent storage of analysis logs and predictions, avoiding external network reliance for its core functions.
[PROMPT_INJECTION]: Analyzes external-controlled data like git commit messages and issue descriptions. While this provides a surface for indirect prompt injection, the skill's purpose is to generate human-readable speculations rather than constructing commands for execution, mitigating the risk.
[CREDENTIALS_UNSAFE]: Incorporates a strict rule ensuring that generated predictions never include sensitive information such as tokens, passwords, or cryptographic keys.
[EXTERNAL_DOWNLOADS]: Mentions tools like ghq and scripts like dig.py, which are treated as local environment dependencies for data discovery.

dream