The Agent Skills Directory

[PROMPT_INJECTION]: A heuristic detection hit was identified in references/practices.md regarding instructions to 'ignore previous rules'. Detailed analysis confirms this is a defensive security guideline advising the agent on how to handle malicious patterns if encountered in the repository's data, rather than an attempt to bypass agent safety filters.
[COMMAND_EXECUTION]: The skill includes a Node.js script (scripts/preflight.mjs) that executes read-only Git commands (git status, git rev-parse, git ls-files) to verify the state of the local filesystem before proposing changes. These commands are hardcoded and used for project safety verification.
[SAFE]: The skill's architecture emphasizes user control and safety. It implements two mandatory 'gates' requiring explicit user confirmation before scanning the project or applying any changes. Furthermore, the instructions explicitly prohibit the movement or modification of sensitive data, such as secrets, credentials, and .env files.

initialize-agentic-memory