skills/xoai/sage/pack-discover/Gen Agent Trust Hub

pack-discover

Pass

Audited by Gen Agent Trust Hub on May 7, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests and processes untrusted data from the codebase and external documentation to generate its output.
  • Ingestion points: Untrusted data enters the agent context via the 'codebase-context' input and user-provided project documentation such as architecture records, API contracts, and style guides.
  • Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions to prevent the agent from obeying instructions potentially embedded within the ingested project context.
  • Capability inventory: The skill performs local file system reads in the 'packs/' directory and has the capability to write the generated brief to '.sage/pack-build/brief.md'.
  • Sanitization: No explicit sanitization, validation, or filtering of the external project context is mentioned before the data is interpolated into the final brief.
Audit Metadata
Risk Level
SAFE
Analyzed
May 7, 2026, 01:31 AM
Security Audit — agent-trust-hub — pack-discover