The Agent Skills Directory

[COMMAND_EXECUTION]: The sample project included in the skill executes local Python scripts using the subprocess module to manage its lifecycle.
fixture/verify.py launches the service.py background process using subprocess.Popen with start_new_session=True.
fixture/tests/test_contract.py uses subprocess.run and subprocess.Popen to execute the adapter's CLI commands during the test suite.
All executions are performed using sys.executable and target specific local files that are part of the skill distribution.
[DYNAMIC_EXECUTION]: The skill implements a strict identity verification mechanism using SHA-256 digests to ensure that the code being executed matches the expected candidate.
fixture/common.py includes functions like candidate_digest() and adapter_digest() that compute cryptographic checksums of the source files.
fixture/verify.py and fixture/product_cli.py validate these digests at runtime, exiting with a stale_or_wrong_target error if the file content does not match the provisioned identity.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill performs local network operations but does not connect to external services.
fixture/common.py and fixture/verify.py use urllib.request to communicate with a local service listening on 127.0.0.1.
The system includes explicit redaction logic in fixture/common.py (evidence_safe_identity) to ensure that ownership tokens and credentials are not stored in the persistent evidence manifests.

verification-adapter