The Agent Skills Directory

[SAFE]: The skill performs expected lifecycle management tasks for a software development environment, such as recording test results and updating state files within the .agent-atelier directory.
[COMMAND_EXECUTION]: The skill utilizes a local script (state-commit) to perform atomic updates to orchestration state. This command is executed via a shell pipe using standard local tools (Bash) and does not involve remote code execution or privilege escalation.
[PROMPT_INJECTION]: The skill processes validation data provided in JSON format. It includes explicit verification steps for work item IDs, candidate branches, and commit hashes, which ensures that only data consistent with the current orchestration state is processed. While it ingests external data (from a validator), the use of structured parsing and precondition checks mitigates potential indirect prompt injection surfaces.

validate