The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/quick_validate_skill.py executes the skill's internal contract test script tests/test_skill_contract.py via subprocess.run. This process is used strictly for local validation of the skill's integrity and file structure.
[DYNAMIC_EXECUTION]: The test script tests/test_skill_contract.py utilizes the compile() function to perform syntax checks on Python files within the repository. This serves as a static code quality check during the skill's self-testing phase.
[INDIRECT_PROMPT_INJECTION]: The skill includes adversarial test fixtures, such as tests/fixtures/adversarial.md, which contain simulated prompt injection attempts. The skill's core instructions in SKILL.md and references/acceptance-tests.md specifically direct the agent to refuse such instructions and maintain objective analysis, serving as a defensive measure rather than a vulnerability.

baseline-ablation-lab