The Agent Skills Directory

[SAFE]: Path Confinement: The lib/profile.mjs module implements strict path validation via confineOutputToPicoclawHome, which prevents path traversal attacks by ensuring all file read/write operations are restricted to the designated Picoclaw home directory.
[SAFE]: Symlink Protection: The scripts/generate_profile.mjs script utilizes O_NOFOLLOW flags and explicit symbolic link checks (isSymbolicLink()) to prevent attackers from using links to gain unauthorized access or write to sensitive files outside the workspace.
[SAFE]: Cryptographic Integrity: The skill uses node:crypto for SHA-256 hashing of files and Ed25519 detached signature verification for supply-chain provenance. This ensures that configuration baselines and release artifacts have not been tampered with.
[SAFE]: Minimal Exposure: The core scripts are designed to be read-only by default and do not perform outbound network requests. Operations are conducted locally using established security state files and local advisory feeds.
[SAFE]: Secure Test Harness: The included regression test (test/picoclaw_security_guardian_sandbox_regression.sh) uses isolated Docker environments to verify installation and execution logic, following secure development and testing principles.

picoclaw-security-guardian