The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of a local Python script scripts/validate_ab_testing_contract.py to validate the structure and content of A/B testing manifests. The script uses standard libraries and does not perform any risky operations like network access or dynamic code evaluation.
[PROMPT_INJECTION]: The skill has a vulnerability surface for indirect prompt injection through its manifest processing workflow.
Ingestion points: The validate_ab_testing_contract.py script and the associated workflow ingest JSON manifest files (e.g., assets/ab-pln-manifest.valid.json) containing natural language fields such as decision_question, interpretation, and follow_up_actions.
Boundary markers: Absent. There are no explicit delimiters or instructions provided to the agent to disregard potential instructions embedded within the manifest fields.
Capability inventory: The skill includes command execution capabilities via a local Python validation script.
Sanitization: The validation script focuses on structural integrity and data types but lacks sanitization or escaping of natural language content to prevent adversarial instruction following.

ab-testing