external-system-debugging

Pass

Audited by Gen Agent Trust Hub on Jul 31, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it guides the agent to ingest and analyze data from untrusted external sources without adequate safety boundaries.\n
  • Ingestion points: Data enters the agent's context through navigator.clipboard.read() output, el.outerHTML DOM captures, and HTTP response data collected via curl as described in SKILL.md and references/template-browser-paste-format.md.\n
  • Boundary markers: The skill does not instruct the agent to use delimiters or 'ignore embedded instructions' markers when processing the gathered external data.\n
  • Capability inventory: The skill enables the agent to recommend and analyze the output of shell commands (e.g., curl, pbpaste) and perform DOM manipulations.\n
  • Sanitization: While the skill suggests avoiding innerHTML in implementation to prevent XSS, it lacks instructions for sanitizing or validating external content before the agent interprets it.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 31, 2026, 01:34 AM
Security Audit — agent-trust-hub — external-system-debugging