From f8c58a52960636f502d60a9e5f34fdc8d30c8212 Mon Sep 17 00:00:00 2001 From: Adryan Eka Vandra Date: Tue, 10 Nov 2026 10:49:52 +0700 Subject: [PATCH] fix(orchestrate): a coordinator ends its turn only with every unblocked unit started or parked Only a child finishing and a PR watch wake an Orchestrate coordinator on T3. A coordinator that closed a round and ended its turn with unblocked units unassigned and nothing running stayed idle until a person wrote: three times in one day, once for 5 h. The anchored rule after "Call no scripts, ..." tells the coordinator to start a worker for every unit that is not blocked, or to record in notes.md why it waits and what will wake it, and to report a turn that leaves work with nothing running as a stop. Fixes #74 --- plugins/pstack/skills/poteto-mode/playbooks/orchestrate.md | 2 +- tools/substitutions.json | 6 ++++++ 2 files changed, 7 insertions(+), 1 deletion(-) diff --git a/plugins/pstack/skills/poteto-mode/playbooks/orchestrate.md b/plugins/pstack/skills/poteto-mode/playbooks/orchestrate.md index c8c4c92..cd25e8f 100644 --- a/plugins/pstack/skills/poteto-mode/playbooks/orchestrate.md +++ b/plugins/pstack/skills/poteto-mode/playbooks/orchestrate.md @@ -2,7 +2,7 @@ **You run the program, never the code. Decide which agent does each piece of work, then check what each one produces.** For a standing project handed to one coordinator chat: multi-day, many stacked PRs, dozens to hundreds of subagents, few human turns. One task driven to a predicate is Autonomous run. One ambitious run that needs a bespoke workflow is figure-it-out. Work that one agent could finish inside the session's budget is not a program. -Run it with your own tools. Spawn workers with `delegate_task`. Keep the program's state in `${XDG_STATE_HOME:-~/.local/state}/t3-pstack/orchestrate//`, with the tracker in its `notes.md`. Watch each PR with `watch_pull_request` when this thread has it. It wakes this thread on failed checks, passed required checks, new comments or reviews, and conflicts. Call no scripts, run no polling loops, and never relay messages or status between agents. +Run it with your own tools. Spawn workers with `delegate_task`. Keep the program's state in `${XDG_STATE_HOME:-~/.local/state}/t3-pstack/orchestrate//`, with the tracker in its `notes.md`. Watch each PR with `watch_pull_request` when this thread has it. It wakes this thread on failed checks, passed required checks, new comments or reviews, and conflicts. Call no scripts, run no polling loops, and never relay messages or status between agents. Only a child finishing and a PR watch wake this thread, so before you end a turn, start a worker for every unit that is not blocked, or write in `notes.md` why it waits and what will wake it. A turn that ends with work left and nothing running or watched stops the program. Report it to the user as a stop, never as a finished round. #### Topology diff --git a/tools/substitutions.json b/tools/substitutions.json index 3081c90..3f29303 100644 --- a/tools/substitutions.json +++ b/tools/substitutions.json @@ -634,6 +634,12 @@ "replacement": "Keep the program's state in `${XDG_STATE_HOME:-~/.local/state}/t3-pstack/orchestrate//`, with the tracker in its `notes.md`. Watch each PR with `watch_pull_request` when this thread has it. It wakes this thread on failed checks, passed required checks, new comments or reviews, and conflicts.", "rationale": "T3 has no Cursor Project tracker or store, so the program's state lives in the port's store directory on local disk, the one Autopilot-full's trail uses. T3's PR and CI event subscription is watch_pull_request, and the wakes it lists are the ones its tool description names. T3 gives a delegated child no watch_pull_request, so a sub-coordinator has none and the condition stays" }, + { + "files": "^skills/poteto-mode/playbooks/orchestrate\\.md$", + "pattern": "Call no scripts, run no polling loops, and never relay messages or status between agents.", + "replacement": "Call no scripts, run no polling loops, and never relay messages or status between agents. Only a child finishing and a PR watch wake this thread, so before you end a turn, start a worker for every unit that is not blocked, or write in `notes.md` why it waits and what will wake it. A turn that ends with work left and nothing running or watched stops the program. Report it to the user as a stop, never as a finished round.", + "rationale": "On T3 a delegate_task child's completion and watch_pull_request are the only things that wake the coordinator's thread. A coordinator that closes a round and ends its turn with unblocked units unassigned and no child or watch running stays idle until a person writes (chhoumann/t3-pstack#74: three times in one day, once for 5 h)" + }, { "files": "^skills/poteto-mode/playbooks/orchestrate\\.md$", "pattern": "Run workers on a new cloud VM (`environment: \"cloud\"`) unless the work needs this machine: a `control-ui` or `control-cli` runtime check (from `cursor-team-kit`), local transcripts, a simulator, or auth that exists only here.",