The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill includes a dedicated configuration resolver script (scripts/config/resolve-config.sh) that specifically checks for and rejects any configuration files containing credential-like keys such as API keys, tokens, or passwords, effectively preventing the accidental exposure of secrets.
[COMMAND_EXECUTION]: Local command execution via shell scripts and Python is restricted to non-mutating Git operations and project-specific diagnostic tools. The implementation uses absolute paths, path traversal checks, and verifies that the target directory is a valid Git repository root.
[DYNAMIC_EXECUTION]: The scripts/run-store.py utility provides owner-only access to historical records using secure file handling techniques, including the use of Unix no-follow primitives and explicit checks for file owner UID and strict access permissions (0700/0600).
[INDIRECT_PROMPT_INJECTION]: The skill is designed with a guided workflow that treats repository data as untrusted. It requires explicit user consultation before making any changes and includes instructions to disregard instructions embedded within data processed by the agent.

woostack-init