The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from repository files and a remote API to generate repository configuration.
Ingestion points: inspect_repo.py scans local project files to detect languages and test commands; load_policy_contract.py fetches JSON policy data from an external API.
Boundary markers: Absent. The instructions direct the agent to treat external data as authoritative without providing explicit prompt delimiters or "ignore" instructions for the interpolated content.
Capability inventory: _skill_cli.py contains a run_command utility using subprocess.run; load_policy_contract.py performs network operations using urllib.request; inspect_repo.py performs file system reads; the agent is instructed to write eddi-ci.yaml to the workspace root.
Sanitization: Absent. The skill does not implement programmatic validation or sanitization of the content fetched from the repository or the remote API before processing.
[EXTERNAL_DOWNLOADS]: The skill fetches a machine-readable policy contract from a remote server.
Evidence: load_policy_contract.py uses urllib.request to fetch JSON from https://cihook.heyeddi.com/api/public/eddi-ci-policy. This URL belongs to the vendor's infrastructure and provides the configuration schema for the generated files.

heyeddi-ci-config