The Agent Skills Directory

[COMMAND_EXECUTION]: The skill mandates the execution of multiple system commands to perform QA duties, including the use of make verify for repository-wide validation, git diff-tree for commit auditing, the gh (GitHub CLI) for Pull Request observation, and a domain-specific tool named roundfix for specification and task-graph checks.
[INDIRECT_PROMPT_INJECTION]: The skill instructions create a surface for indirect prompt injection by requiring the agent to process data from various untrusted sources that could contain malicious instructions designed to influence the agent's behavior during a QA run.
Ingestion points: The agent is instructed in SKILL.md to read project-level files such as _prd.md and task_NN.md, monitor Pull Request data via gh, and exercise public application interfaces (frontend DOM, API responses, and CLI outputs).
Boundary markers: The instructions do not define technical delimiters or specific guardrail prompts to isolate the data being tested from the agent's core instructions.
Capability inventory: The skill provides the agent with the authority to execute shell commands (make, git, gh, roundfix), automate browser-based user flows, and perform local file system writes to document QA findings.
Sanitization: There are no explicit requirements or steps for the agent to sanitize, escape, or validate the content ingested from the application or its documentation before using that information to drive logic or generate reports.

qa-gate