harness-engineering

Pass

Audited by Gen Agent Trust Hub on Aug 8, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze untrusted data from repository files such as AGENTS.md, CI workflows, and documentation. This creates a potential surface for indirect prompt injection where instructions embedded in these files could influence agent behavior.
  • Ingestion points: The agent is directed to inspect AGENTS.md, contributor docs, CI workflows, and repository scripts in SKILL.md (Workflow Step 1).
  • Boundary markers: There are no explicit instructions or delimiters provided to help the agent distinguish between static data and potentially malicious instructions within the analyzed files.
  • Capability inventory: The skill workflow involves creating or editing repository files and defining new CI guardrails, which implies the agent will use file-system write and potentially shell execution capabilities (SKILL.md, Deliverable format).
  • Sanitization: No sanitization or validation logic is defined for the content extracted from repository files.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 8, 2026, 06:23 PM
Security Audit — agent-trust-hub — harness-engineering