The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a Python Zip Application (cut.pyz) located at skills/cut/scripts/cut.pyz. This tool is used for all "gate" operations, including contract resolution, baseline freezing, and evidence issuance. The skill explicitly requires resolving this command as a specific file path to prevent execution of unauthorized binaries from the system PATH.
[DYNAMIC_EXECUTION]: The skill relies on a .pyz file, which is an executable archive of Python code. This allows for the execution of complex logic that is bundled with the skill rather than being written directly in the markdown instructions.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data from specifications and user-provided reproductions to define execution parameters (argv and cwd).
Ingestion points: External inputs are processed via the spec_ref and reproduction arguments in SKILL.md and references/gate-workflow.md to generate test plans.
Boundary markers: The instructions do not specify explicit boundary markers (like XML tags or specific delimiters) for the ingested data.
Capability inventory: The skill possesses the capability to execute arbitrary subprocesses through the red-gate tool and standard Python test runners (e.g., pytest, unittest).
Sanitization: The skill mitigates potential injection by requiring that all commands be passed as argv arrays and strictly forbidding shell evaluation (never shell evaluation).

cut