The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from potentially untrusted external sources such as PRDs, design documents, and tickets to generate task files. This creates a surface for indirect prompt injection if those sources contain malicious instructions designed to influence the agent's planning output. The skill includes mitigations by requiring all claims to be grounded in existing repository code and conventions.
[EXTERNAL_DOWNLOADS]: The instructions direct the agent to fetch design links and artifact URLs referenced in source materials to transcribe decisions into task criteria. While this involves network operations to external domains, it is a functional requirement for the planning process.
[COMMAND_EXECUTION]: The skill suggests using the 'project's documentation tools' to resolve library-specific facts that are not established locally. This involves executing local CLI utilities within the agent's authorized development environment.

wtk-plan