doc: document offset file semantics

offset points to chat position after the last user prompt;
tail -c +$((offset+1)) yields the model response for that turn.
Replace stale reply references in prompts, help, ideas, and skills.
Document offset usage in concurrent sessions pattern and shell script.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

doc/IDEAS.md

@@ -34,6 +34,8 @@ See `prompts/08_subagents.md` and `skills/session-management/SKILL.md` for usage
 
 For workflows where multiple long-lived agents work in parallel and need to coordinate, the mechanism is direct prompt passing. Each agent is told its own session ID and its peers' IDs at startup. When an agent finishes a step and wants to hand off, it writes the result to another session's `prompt` — a plain file write from within `execute_code`. Writing to `prompt` queues by default, so a busy agent will not drop a message.
 
+When a peer needs to read back the response it triggered, it polls the target session's `state` until idle, then reads `offset` and extracts the new output: `tail -c +$((offset+1)) s/$peer/chat`. The server sets `offset` to the byte position immediately after the user prompt is appended, so everything from that position to EOF is exactly the model response for that turn.
+
 No external coordinator is needed after the initial kickoff. The agents themselves route work based on context in their prompts. A human can intervene at any time by writing to any session's `prompt`.
 
 #### Example: develop → review → test
@@ -70,7 +72,7 @@ Each agent can use a different backend, model, or tool configuration. The bounda
 
 #### Predefined workflows
 
-When the coordination logic is known in advance, a shell script can drive the entire workflow externally. Named sessions make routing unambiguous. The script captures a chat offset before each prompt submission, waits for the agent to finish, and reads only the new output — no `reply` file needed.
+When the coordination logic is known in advance, a shell script can drive the entire workflow externally. Named sessions make routing unambiguous. The server updates `offset` to the byte position immediately after the user prompt is appended to chat. After the agent goes idle, `tail -c +$((offset+1))` yields exactly the model response for that turn.
 
 ```sh
 #!/bin/sh
@@ -81,14 +83,15 @@ echo "name=developer agent=developer cwd=$cwd" > s/new
 echo "name=reviewer  agent=reviewer  cwd=$cwd" > s/new
 echo "name=tester    agent=tester    cwd=$cwd" > s/new
 
-# Send a prompt to a session and return only the new output.
+# Send a prompt and return the model response.
+# The server sets offset after the user prompt is appended to chat;
+# everything from offset to EOF is the model response for that turn.
 send() {
     sid=$1; shift
     printf '%s' "$*" > s/$sid/prompt
     while [ "$(cat s/$sid/state)" = "idle" ]; do sleep 0.2; done
     while [ "$(cat s/$sid/state)" != "idle" ]; do sleep 1; done
-    offset=$(cat s/$sid/offset)
-    tail -c +$((offset + 1)) s/$sid/chat
+    tail -c +$(($(cat s/$sid/offset) + 1)) s/$sid/chat
 }
 
 reply=$(send developer "Implement a function that parses a JSON config file.")

doc/help.md

@@ -40,7 +40,7 @@ ollie/
       agent       active agent (r/w)
       backend     active backend (r/w)
       chat        conversation log (ro)
-      offset      byte offset of chat EOF before last prompt submit (ro)
+      offset      byte offset of chat after last user prompt; tail -c +$((offset+1)) gives model response (ro)
       ctl         control (w)                stop, kill, clear, compact, rn <name>
       ctxsz       context size (ro)
       cwd         working directory (r/w)

prompts/01_ollie.md

@@ -30,7 +30,7 @@ Each session is a directory `{{.OllieMountPath}}/s/{session-id}/`. Your current
 | `{{.OllieMountPath}}/s/{id}/enqueue` | write | Queue a prompt for execution after the current turn |
 | `{{.OllieMountPath}}/s/{id}/dequeue` | read  | Pop the next queued prompt |
 | `{{.OllieMountPath}}/s/{id}/chat`    | read  | Full cumulative chat history (tail-able) |
-| `{{.OllieMountPath}}/s/{id}/reply`   | read  | Assistant text from the most recent turn (tail-able) |
+| `{{.OllieMountPath}}/s/{id}/offset`  | read  | Byte offset of chat immediately after the last user prompt; `tail -c +$((offset+1))` yields the model response for that turn |
 | `{{.OllieMountPath}}/s/{id}/state`   | read  | Current agent state: `idle`, `thinking`, or `calling: <tool>` |
 | `{{.OllieMountPath}}/s/{id}/backend` | r/w  | Active backend name |
 | `{{.OllieMountPath}}/s/{id}/agent`   | r/w  | Active agent name |

skills

@@ -1 +1 @@
-Subproject commit f48f2086824a27530872748e62ed1fa748360fcf
+Subproject commit 2dd6fea2e63db4f74cb61d7b66eb7b28334d8bc9