aitask-pickrem

Pick and implement a task in remote/non-interactive mode. All decisions from execution profile - no AskUserQuestion calls.

SKILL.md

GitHub

This is a profile-aware skill stub. Pre-rendered variants for headless profiles (currently remote) are committed to the repo so the skill works in environments where the rendering toolchain (minijinja) is unavailable — e.g., Claude Code Web. Other profiles render on demand. Execute these steps in order, then stop:

  1. Resolve active profile. Parse ARGUMENTS for --profile <name>. If found, use that as <profile> and remove the --profile <name> pair from ARGUMENTS. Otherwise run: ./.aitask-scripts/aitask_skill_resolve_profile.sh pickrem and use the single-line stdout as <profile>.

  2. Render only if needed. If the committed pre-rendered file at .claude/skills/aitask-pickrem-<profile>-/SKILL.md already exists, skip this step. Otherwise run: ./.aitask-scripts/aitask_skill_render.sh aitask-pickrem --profile <profile> --agent claude (requires ait setup to have installed minijinja).

  3. Dispatch via Read-and-follow. Read the file at .claude/skills/aitask-pickrem-<profile>-/SKILL.md and execute its instructions as if they were this skill, forwarding the (possibly stripped) ARGUMENTS unchanged.