miro-debug-bundle

Pass

Audited by Gen Agent Trust Hub on Sep 10, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external log files located at /var/log/app, ~/.npm/_logs, and other application directories. If these logs contain malicious instructions from an attacker (e.g., via failed login attempts or specific request patterns), the agent could potentially process them as instructions.
  • Ingestion points: Step 4 in SKILL.md searches log directories for 'miro' related entries.
  • Boundary markers: No specific delimiters or instructions to ignore embedded content are present for the log aggregation step.
  • Capability inventory: The skill utilizes curl for network requests and Bash for command execution, creating a potential surface for exploitation if an injection is successful.
  • Sanitization: The script uses sed to redact Bearer tokens and masks environment variable values in .env files, but it does not filter the logs for natural language command patterns.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 10, 2026, 11:34 PM
Security Audit — agent-trust-hub — miro-debug-bundle