The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python scripts (seal_document.py, audit_document.py) using subprocess.run within its protocol runtime. These executions are scoped to the skill's own internal logic and the target repository provided by the user. The commands are constructed from templates defined in skill-protocol.json and use strict shell=False execution.
[DYNAMIC_EXECUTION]: The protocol runtime (scripts/_skill_protocol_runtime.py) utilizes importlib.import_module to load its local diagnostic contract. This is a standard library pattern used for modularity within the skill's own package structure and does not provide an external injection vector.
[PATH_SANITIZATION]: The skill implements significant safety checks in scripts/bundle_workspace.py and scripts/_skill_protocol_runtime.py to ensure that execution and temporary state storage occur outside the installed skill directory and that audit reports are correctly ignored by Git, preventing accidental state contamination or repository pollution.
[SAFE]: All findings are associated with the intended primary purpose of the skill (document validation and auditing). The tool uses a 'least privilege' approach by requiring sensitive state (the document bundle) to be stored in a temporary directory outside the target repository.

scribe