heyeddi-ci-config
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches living policy configuration (knobs, rules, and guides) from the vendor's API at https://cihook.heyeddi.com/api/public/eddi-ci-policy to ensure configuration validity.
- [DYNAMIC_EXECUTION]: The script load_policy_contract.py dynamically modifies sys.path to import modules from computed local paths if a vendor repository is detected, which allows it to interface with local development environments.
- [INDIRECT_PROMPT_INJECTION]: The skill treats externally fetched JSON as authoritative instructions for the agent's behavior. 1. Ingestion points: load_policy_contract.py fetches data from the vendor's public API. 2. Boundary markers: No delimiters or protective warnings are applied to the ingested instructions. 3. Capability inventory: The agent is instructed to write eddi-ci.yaml based on this data. 4. Sanitization: The remote content is used directly without validation or escaping.
- [COMMAND_EXECUTION]: Shared helper scripts include infrastructure for executing shell commands via subprocess.run. While primarily used for diagnostics or future extension, this provides a capability for command execution within the skill's scripts environment.
Audit Metadata