The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to execute local build and test commands (e.g., npm run test, pytest, go test).
Evidence: Found in scripts/discover_and_verify.py and scripts/_skill_cli.py. The execution is limited to verified repository commands discovered in files like package.json, Makefile, or Cargo.toml. The gh CLI is also used to fetch read-only PR information in scripts/fetch_failing_checks.py.
[PRIVILEGE_ESCALATION]: The skill includes an explicit safety mechanism to prevent the agent from merging PRs without manual user authorization.
Evidence: scripts/assert_no_merge.py acts as a hard gate, scanning for merge-like strings and requiring an "authorize merge" token from the user before allowing the agent to proceed with destructive actions.
[DATA_EXPOSURE]: The skill writes local reports and raw JSON data containing CI failure logs to a .heyeddi/docs/ directory.
Evidence: Seen in scripts/write_ci_fails_report.py. The instructions and ANTI_PATTERNS.md explicitly warn the agent not to commit these ephemeral artifacts to the repository.

heyeddi-ci-fails