omni-cli-tools
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill exposes endpoints for retrieving sensitive information, such as API keys via
GET /api/cli-tools/keysand configuration backups viaGET /api/cli-tools/backups. - [DATA_EXFILTRATION]: Instructions specify reading from local sensitive configuration paths, including
~/.codewhale/config.tomland legacy fallback~/.deepseek/config.toml. - [INDIRECT_PROMPT_INJECTION]: The skill establishes a high-capability attack surface for indirect prompt injection.
- Ingestion points: The agent ingests information about CLI tools and user-provided configuration values which are then sent to management endpoints in
references/endpoints.md. - Boundary markers: None present in the API request templates or instructions.
- Capability inventory: The skill allows arbitrary modification of tool configurations, including
apiKey,baseUrl, and proxy settings (antigravity-mitm) viaPOSTandPUTrequests inreferences/endpoints.md. - Sanitization: No explicit sanitization or validation of the values sent to the API is documented within the skill.
Audit Metadata