The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external task descriptions (".tasks" files) and design specifications, which introduces a surface where malicious instructions could be embedded in source data. The skill mitigates this risk by requiring that all work be grounded in existing project conventions, providing a "refuse gate" for ambiguous or open-ended task descriptions, and establishing a mandatory verification process performed by a fresh sub-agent (Verifier) that lacks context from the implementation phase.
[COMMAND_EXECUTION]: The skill instructs the agent to execute standard software development and testing tools such as git, pytest, jest, and bun. These operations are essential to the skill's purpose of code implementation and verification. The instructions emphasize using the project's existing commands and CI configurations rather than arbitrary or external scripts, and explicitly state that the skill does not install additional software.

wtk-implement