The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands including cat, jq, git, and npm to analyze project structure, read configuration files, and execute test suites. These actions are aligned with the skill's stated purpose of quality assurance within a development workflow.
[PROMPT_INJECTION]: The workflow involves reading content from prd.json and passing it to a test-creator sub-agent. This pattern represents an indirect prompt injection vector where instructions embedded inside the prd.json file could potentially influence the sub-agent's behavior.
Ingestion points: Data is read from prd.json using cat and jq in SKILL.md.
Boundary markers: The prompt template for the test-creator does not employ specific delimiters or "ignore previous instructions" warnings to isolate the untrusted input from the logic.
Capability inventory: The skill environment permits file system access, shell command execution via npm, and sub-agent task delegation as seen in the Task call.
Sanitization: There is no evidence of input validation or escaping for the data fields (such as acceptance criteria or GDD sections) extracted from the PRD before interpolation into the sub-agent prompt.

qa-test-creation