setup-harness
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from local configuration files which could potentially contain malicious instructions.
- Ingestion points: Existing configuration files such as
~/.codex/AGENTS.md,~/.claude/CLAUDE.md, and Cursor's User Rules interface. - Boundary markers: The agent is explicitly instructed to treat retrieved content as evidence, not instructions, and to preserve unrelated content during the merge.
- Capability inventory: The skill is restricted to file-editing access; network access and external tool calls are disabled via the
disable-model-invocation: truefrontmatter setting. - Sanitization: The skill requires an explicit human review and approval of a complete diff or proposal before any changes are written to the filesystem.
- [SAFE]: The skill's operations are confined to local configuration management of well-known AI tools. No network exfiltration, obfuscation, or unauthorized command execution patterns were detected. References point to official documentation from trusted vendors.
Audit Metadata