The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/load_policy_contract.py performs dynamic code loading by modifying sys.path to import a module from relative directory paths (e.g., heyeddi-ci/backend/app/policy_insights.py) computed at runtime based on the project root or current working directory.
[DYNAMIC_EXECUTION]: The script scripts/assert_runners_placeholder.py uses runpy.run_path to dynamically execute a separate Python script (assert_runners_claims.py) located in the same directory.
[INDIRECT_PROMPT_INJECTION]: The inspect_repo tool scans and reads the contents of files within the project root to detect programming languages and build/test commands, creating a surface for indirect prompt injection if repo contents are maliciously crafted.
Ingestion points: scripts/inspect_repo.py recursively reads files in the repository using rglob and read_text.
Boundary markers: Skill instructions emphasize that the agent must use evidenced commands and not invent execution claims, though no explicit delimiter boundaries are defined in the scripts.
Capability inventory: The skill possesses the capability to fetch remote data (load_policy_contract) and perform dynamic code execution via imports.
Sanitization: The tool uses regex to scan for specific patterns rather than executing the ingested data directly.
[EXTERNAL_DOWNLOADS]: scripts/load_policy_contract.py fetches a machine-readable policy contract from https://cihook.heyeddi.com/api/public/eddi-ci-policy, which is the official vendor domain for the CI service.

heyeddi-ci-runners