build-agent
Pass
Audited by Gen Agent Trust Hub on Jul 7, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: Credential management. The skill facilitates the use of the Agenta API by instructing users to store their API key in a local .env file or environment variables, while explicitly recommending the use of .gitignore to prevent accidental credential leakage.
- [COMMAND_EXECUTION]: API interactions. Bundled shell scripts use curl and jq to interact with the official Agenta API (agenta.ai). JSON payloads are constructed safely using jq arguments, which protects against command injection vulnerabilities.
- [EXTERNAL_DOWNLOADS]: System preflight. The check-prereqs.sh script verifies the presence of necessary system tools (bash, curl, jq) and provides standard installation instructions for various platforms without performing any unauthorized automated downloads or executions.
- [PROMPT_INJECTION]: Indirect injection surface. The skill allows building agents that process external data, creating an indirect prompt injection surface. * Ingestion points: External data from event attributes is mapped into agent inputs (e.g., in references/trigger-inputs.md). * Boundary markers: The skill provides documentation in references/writing-instructions.md advising users to use explicit, numbered procedures to constrain agent behavior. * Capability inventory: Built agents may have access to integrated tools (e.g., GitHub, Slack) as defined in the provided configuration schemas. * Sanitization: No specific sanitization logic is implemented for external event data being interpolated into the prompts.
Audit Metadata