The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local Bash scripts (init-wiki.sh, check-wiki.sh, archive-journal.sh) to perform maintenance tasks. These scripts are implemented with security best practices, such as using set -euo pipefail for robust error handling, mktemp for safe temporary file management, and using -- to prevent positional argument injection in file operations like cp, mv, and rm. All operations are confined to the local filesystem within the user-specified wiki directory.
[INDIRECT_PROMPT_INJECTION]: As a tool for managing user-supplied markdown content, the skill inherently possesses an indirect prompt injection surface. If a user's notes contain malicious instructions, an agent reading those notes might be influenced. However, the skill minimizes this risk by using deterministic scripts for organizational logic (e.g., parsing dates, checking links) rather than having the LLM interpret and execute the content directly. The documentation emphasizes reading on demand, which limits the exposure of potentially malicious content to only when it is explicitly requested.
[SAFE]: No network activity, obfuscation, credential exposure, or persistence mechanisms were detected. The skill's behavior matches its stated purpose of managing a structured knowledge base efficiently. The provided scripts are part of the skill package and do not perform any remote code execution or unauthorized data exfiltration.

second-brain