The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including tickets, feedback, and knowledge-base documents. This introduces a surface for indirect prompt injection. The skill mitigates this by including a dedicated security policy in references/UNTRUSTED_CONTENT.md, which instructs the agent to treat such content as data rather than instructions and mandates user confirmation for all write operations.
[COMMAND_EXECUTION]: The skill executes multiple commands via the cawplan CLI tool to manage requirement resolution, data fetching, and the test-point review workflow. These operations are essential to the skill's primary purpose and are performed through the authorized Bash tool.
[EXTERNAL_DOWNLOADS]: The skill fetches requirement documentation and existing test points from the CawPlan platform. These downloads are performed using vendor-provided tools to synchronize the local state with the remote system.

cawplan-testpoint-generate