production-readiness

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a framework for collecting and evaluating production evidence (e.g., security reviews, dependency maps, and SLO declarations). These evidence summaries are provided as external input, creating a surface where maliciously crafted text could attempt to manipulate the agent's launch decision.
  • Ingestion points: The "Source / evidence" fields within the production evidence checklist in SKILL.md and the references/readiness-record.md template.
  • Boundary markers: Absent; there are no specific instructions to the agent to treat evidence text as untrusted or to ignore embedded commands.
  • Capability inventory: The skill possesses no dangerous capabilities; it has no subprocess calls, file-writing tools, or network access.
  • Sanitization: Absent; the skill relies on the agent's internal reasoning to parse the provided text.
  • [NO_CODE]: The skill consists solely of Markdown instructions, templates, and JSON evaluation data. It does not ship with scripts, binaries, or any form of executable code.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 05:57 PM
Security Audit — agent-trust-hub — production-readiness