From a713760f27b2ced1f5e076fc5abe862e66311927 Mon Sep 17 00:00:00 2001 From: Levi Neely Date: Fri, 10 Apr 2026 02:31:18 +0200 Subject: [PATCH] =?UTF-8?q?Drop=20"orchestrator"=20=E2=80=94=20it's=20just?= =?UTF-8?q?=20a=20shell=20script?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Co-Authored-By: Claude Sonnet 4.6 --- README.md | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index 05c1f6b..f0fb52f 100644 --- a/README.md +++ b/README.md @@ -137,11 +137,11 @@ echo "kill " > ~/mnt/ollie/ctl Each session exposes a `reply` file containing only the assistant text from the most recently completed turn. Writing to `prompt` clears `reply` as a side-effect. Together these make agent-to-agent handoffs expressible as plain file operations — no message queue, no orchestration framework, no shared memory. -A shell script is the orchestrator. The agents communicate only through what the script explicitly passes between them, so information flow is fully under your control. Each agent can use a different backend, model, or tool configuration. A human can intervene at any point by writing directly to a session's `prompt`. +A shell script moves text between sessions. The agents communicate only through what the script explicitly passes between them, so information flow is fully under your control. Each agent can use a different backend, model, or tool configuration. A human can intervene at any point by writing directly to a session's `prompt`. #### Example: develop → review → test -Three sessions with specialized agent configs run a feedback loop. The reviewer ends its response with `LGTM` (proceed) or `PTAL` (revise). The tester ends with `Approved` (done) or `Rejected` (revise). The orchestrator parses the verdict with `grep` and routes accordingly. +Three sessions with specialized agent configs run a feedback loop. The reviewer ends its response with `LGTM` (proceed) or `PTAL` (revise). The tester ends with `Approved` (done) or `Rejected` (revise). A shell script reads the verdict with `grep` and routes accordingly. ```sh #!/bin/sh @@ -194,7 +194,7 @@ while true; do done ``` -Each agent only sees what the orchestrator explicitly sends it. Each can use a different backend, model, or tool configuration. A human can intervene at any point by writing directly to a session's `prompt`. The boundary between fully automated and human-in-the-loop is just whether the script pauses to ask. +Each agent only sees what the script explicitly sends it. Each can use a different backend, model, or tool configuration. A human can intervene at any point by writing directly to a session's `prompt`. The boundary between fully automated and human-in-the-loop is just whether the script pauses to ask. ## Example shell session