The Agent Skills Directory

[SAFE]: The skill defines a structured methodology for maintaining reliable automated tests, focusing on legitimate development workflows.
[PROMPT_INJECTION]: The skill is designed to ingest external data from project context files (.agents/qa-project-context.md) and CI failure logs. This surface for indirect prompt injection is mitigated by the skill's built-in 'Observable Repair Workflow' which requires high confidence scores for automated changes and includes 'Intent Fidelity Checks' to ensure the test purpose is not subverted.
[COMMAND_EXECUTION]: The skill utilizes standard software testing and version control CLI tools (such as playwright, jest, pytest, and git) to diagnose failures and verify fixes. These commands are used for their intended primary purposes within a QA context.
[SAFE]: The self-healing mechanisms described (automatically repairing locators) are limited to modifying local test scripts and include rollback procedures, aligning with the skill's stated purpose of test maintenance.

test-reliability