The Agent Skills Directory

[SAFE]: The skill implements a robust reasoning process and a local data storage system. The 'Truth Vault' logic in references/truth_vault.py includes comprehensive security checks to prevent path traversal and ensure file operations remain within the designated project directory.
[COMMAND_EXECUTION]: The file references/truth_vault_security_tests.py uses subprocess.run() to execute the local truth_vault.py script for regression testing. This usage is safe as it is limited to the test environment and executes a known local script using a resolved path.
[DATA_EXFILTRATION]: No exfiltration vectors were found. The Truth Vault utility manages local JSONL files and does not perform network operations or access sensitive system paths outside its defined scope.
[PROMPT_INJECTION]: The skill uses structured 'Phase' gates and 'HARD-GATE' instructions to enforce reasoning discipline. These are process-oriented and do not contain patterns typical of malicious prompt injection or behavior overrides.

first-principles-thinking