create-agent
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation of AI agents by interpolating user-provided instructions and constraints into a generated YAML specification. This architecture creates an attack surface where untrusted user input could lead to prompt injection in the resulting agent.
- Ingestion points: Phase 2 requirements gathering (Goal, Inputs, Workflow, Constraints, Done criteria).
- Boundary markers: No mandatory structural delimiters or 'ignore' instructions are enforced for the user-supplied content in the final spec, although a ## RULES section is recommended.
- Capability inventory: The skill manifest agents using harness_create and harness_update tools from the harness-mcp-v2 server and references a container image from pkg.harness.io.
- Sanitization: Security depends on the human-in-the-loop review process defined in Phase 4 before resource creation.
- [EXTERNAL_DOWNLOADS]: The generated agent specification references a vendor-provided container image: pkg.harness.io/vrvdt5ius7uwygso8s0bia/harness-agents/harness-ai-agent:latest.
Audit Metadata