infrastructure-management

Pass

Audited by Gen Agent Trust Hub on Aug 13, 2026

Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: The skill includes a metadata file (skill-report.json) that contains pre-populated security audit results declaring the skill as "safe". This is a self-referential content pattern designed to influence automated security analyzers and potentially bypass independent evaluation.
  • Ingestion points: The skill-report.json file is processed as part of the skill's metadata and context.
  • Boundary markers: Absent; the safety claims are presented as authoritative metadata without delimiters.
  • Capability inventory: None; the skill contains no executable scripts or tools, limiting the potential impact of an injection.
  • Sanitization: Absent; the skill lacks mechanisms to isolate its own metadata claims from the analysis environment.
  • [NO_CODE]: The skill consists solely of Markdown documentation (SKILL.md) and JSON metadata (skill-report.json). No executable code, shell scripts, or binary dependencies are included, which reduces the overall risk profile despite the metadata poisoning attempt.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 13, 2026, 05:08 PM
Security Audit — agent-trust-hub — infrastructure-management