For your agent / Page 49 of 70
This page is one part of The Ishoo manual. The complete manual is a single page, short enough to read in one go: ishoo.app/docs/
For your agent
An AI agent uses Ishoo in a loop: ishoo_status for the plan and who holds what, ishoo_claim to take an issue and get a worktree, normal work, then ishoo_land with a note of what changed and why and the tests it ran. Ishoo trusts the agent; it nudges and rarely refuses.
These are the exact instructions an agent receives when it connects:
Ishoo coordinates agents and keeps a trace of what landed. Loop: ishoo_status (the why and who is on what), ishoo_claim (get a worktree), work as you normally would, ishoo_land (note what and why; name tests you ran). Ishoo trusts you: it nudges, it rarely refuses. Keep the plan's outcome in view; pick work by what moves it most. Keep the record for the person: when they settle how something should work and why, offer once to record it as a decision; when they say they want something you aren't doing now, offer once to file it as an issue (or a plan, for a larger outcome). Offer, don't file unasked. Skip passing remarks, routine implementation choices, and anything already filed (ishoo_find first).
The project's CLAUDE.md or AGENTS.md carries one line to the same effect. Every tool is on its own page in this chapter, with each parameter.