root-cause-analysis

Pass

Audited by Gen Agent Trust Hub on Mar 24, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is designed to analyze potentially untrusted external data, creating a surface for indirect prompt injection.\n
  • Ingestion points: Systematically defined in SKILL.md (Phase 1: Gather Before You Reason) and throughout the reference files which process stack traces and logs.\n
  • Boundary markers: The skill employs XML-style tags such as <gathering>, <reading_errors>, and <hypothesis> to delimit diagnostic phases.\n
  • Capability inventory: The methodology encourages the agent to use powerful system tools including strace, lsof, journalctl, and cloud CLI utilities like aws and docker.\n
  • Sanitization: No specific data sanitization or filtering of the ingested log/error content is specified in the instructions.\n- [COMMAND_EXECUTION]: The skill provides comprehensive instructions for executing shell commands to perform system and network diagnostics.\n
  • Evidence: references/platform-linux.md suggests the use of strace, ss, ip, and curl for system inspection. references/platform-aws.md details various AWS CLI commands for permission and service troubleshooting. references/platform-docker.md includes commands for inspecting container state and networking.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 24, 2026, 04:31 PM
Security Audit — agent-trust-hub — root-cause-analysis