The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes node .agents/pstack/lead.mjs to retrieve current session metadata such as the active model and effort level. It also utilizes git grep to search the local plans directory for draft files associated with specific links or paths.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from previous plans and pull request (PR) bodies as part of the 'Redo' workflow. While these inputs constitute an attack surface for indirect prompt injection, the skill includes a specific safety instruction to avoid reading published pages to prevent ingesting content that might be owned or modified by unauthorized accounts.

redo