proactive-agent
Pass
Audited by Gen Agent Trust Hub on Mar 21, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill implements a comprehensive defense-in-depth strategy against prompt injection. It includes explicit instructions to treat external data (emails, websites, PDFs) as data rather than commands and provides a heartbeat system to scan for injection markers like 'ignore previous instructions'.
- [COMMAND_EXECUTION]: The skill provides a shell script (
scripts/security-audit.sh) for verifying local system security and encourages the use of CLI tools for 'self-healing' and autonomous problem-solving. These capabilities are constrained by the 'Self-Improvement Guardrails' (ADL/VFM protocols) described in the core architecture. - [DATA_EXFILTRATION]: Diagnostic routines access local sensitive paths, including
~/.clawdbot/clawdbot.json,.credentials/, and local logs, to perform security audits and verify implementation integrity. These operations are local and associated with the skill's primary defensive purpose. - [SAFE]: The architecture includes advanced alignment systems and context survival protocols (WAL Protocol) that enhance the agent's reliability and security posture by ensuring critical decisions are captured before context loss.
Audit Metadata