The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill accesses local configuration and state files (bkit.config.json, .bkit/state/trust-profile.json) and writes to a runtime log file. All operations are restricted to the project directory, and no network requests or sensitive credential access were identified.- [INDIRECT_PROMPT_INJECTION]: The skill processes a user-supplied feature name argument. However, it incorporates specific mitigations by using templates from a local configuration file and enforcing path rooting to prevent directory traversal. The audit log implementation also includes sanitization for user-controlled data to maintain integrity.- [DYNAMIC_EXECUTION]: The skill integrates with a local JavaScript module (lib/control/fast-track.js) for its core logic. It does not perform runtime code generation, remote script execution, or unsafe deserialization of external data.- [COMMAND_EXECUTION]: The skill defines automation levels and checkpoint bypasses, but these refer to the internal workflow logic rather than system-level command execution or privilege escalation.

pdca-fast-track