The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local controller binary named resolve-c3 and various internal Python scripts to perform preflight checks, validation, and material workflow tasks. These executions are scoped to the skill's primary function of code resolution and conform to standard tool usage patterns.
[SAFE]: The skill employs a 'fail-closed' design. If the required controller capability is missing, dangerous operations such as delivery mutation, committing, and pushing are explicitly forbidden and blocked by the preflight logic.
[SAFE]: Validation tools within the skill (e.g., acceptance_contract_gate.py, mbkc_gate.py) use secure data handling practices, such as yaml.safe_load, to process configuration and state files.
[SAFE]: The instructions establish a rigorous 'Intent-Closed' protocol that prevents the agent from making unauthorized modifications based on raw prose. All changes must be derived from a 'Sealed Acceptance Contract' and a 'Counterexample Basis', which significantly reduces the risk of accidental or malicious code injection during the review process.

resolve