The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves analyzing untrusted repository content, creating a significant attack surface for indirect prompt injection.
Ingestion points: Repository manifests (e.g., package.json), documentation (e.g., README.md), CI configurations, and test source code are read into the agent's context by the inspect_test_health.py script and the SKILL.md workflow.
Boundary markers: The instructions include a manual mitigation step ("Inspect commands before executing them"), but the skill lacks automated delimiters or sanitization for the ingested repository data.
Capability inventory: The skill uses a bundled Python helper to perform filesystem and Git analysis and encourages the agent to execute shell commands found in the repository.
Sanitization: No automated sanitization or filtering is applied to the project content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill identifies and recommends the execution of test commands discovered in the target repository's manifests (e.g., npm test, pytest, Makefile targets). This behavior involves executing code defined by third parties, which may be malicious if the project being audited is untrustworthy.

test-health