The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The load_policy_contract tool fetches a JSON policy contract from cihook.heyeddi.com. The script allows this URL to be overridden via the url parameter or HEYEDDI_CI_POLICY_URL environment variable, enabling downloads from arbitrary remote sources.
[REMOTE_CODE_EXECUTION]: The script scripts/load_policy_contract.py implements dynamic loading of Python modules. It inserts a path derived from the HEYEDDI_CI_ROOT environment variable or the --heyeddi-ci-root argument into sys.path and performs an import (from app.policy_insights import public_policy_contract), which allows execution of arbitrary local code.
[COMMAND_EXECUTION]: The utility script scripts/_skill_cli.py includes a run_command function that utilizes subprocess.run to execute list-based commands. While this utility is not directly exposed as a primary tool, its presence provides a capability surface for command execution within the skill's operational environment.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it is instructed to treat external JSON data as authoritative guidance for its behavior.
Ingestion points: Data is ingested from the remote endpoint https://cihook.heyeddi.com/api/public/eddi-ci-policy via the load_policy_contract tool.
Boundary markers: The instructions lack explicit delimiters or warnings to treat the fetched guide and rules fields as untrusted content.
Capability inventory: The skill possesses capabilities for network requests (load_policy_contract), repository scanning (inspect_repo), and file system writes (eddi-ci.yaml).
Sanitization: The script validates that the response is JSON but does not sanitize the natural language strings within the response that are used to influence agent decisions.

heyeddi-ci-config