The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local scripts including scripts/init.sh for environment scaffolding and check_acceptance.py for goal verification. These tools are identified as components of the agent-completion-gate ecosystem associated with the author.
[DATA_EXPOSURE]: Processes and writes state and configuration to local files such as surface_inventory.yaml, acceptance_manifest.yaml, and completion_candidate.yaml to track task progress.
[PROMPT_INJECTION]: While the skill ingests natural language goals from users, it implements a mandatory human-in-the-loop confirmation step (Step 3) where acceptance criteria must be explicitly approved before any implementation work begins, mitigating accidental obedience to malicious instructions embedded in goals.

goal-compile