Document runtime integration for /ac command

README.md

@@ -73,16 +73,31 @@ cp -R openclaw-auto-continue ~/.openclaw/workspace/skills/
 - destructive action confirmation
 - user decision points
 
+### 如果其他電腦出現 `/ac` unknown
+這通常代表目標機器的 **live OpenClaw runtime 還沒有真正實作這個 command**。
+
+所以這個 repo 不只要教怎麼用，還要教怎麼把 `/ac` 接進 runtime。
+請看：
+
+- `openclaw-auto-continue/references/runtime-integration.md`
+
+這份文件會說明：
+- runtime handler 應該做什麼
+- `/ac status`、`/ac on`、`/ac off` 應該如何改變 session
+- 如何驗證另一台真實機器
+- 如果 skill 存在但 command 不存在，應該怎麼排查
+
 ### 給 Agent 看
 如果 OpenClaw agent 要判斷怎麼使用這個 repo：
 
-1. 讀 `openclaw-auto-continue/SKILL.md`
+1. 先讀 `openclaw-auto-continue/SKILL.md`
-2. 依照裡面的 trigger conditions 與 command usage 使用
+2. 如果目標機器不認得 `/ac`，再讀 `openclaw-auto-continue/references/runtime-integration.md`
 3. 在對應 chat / session 中使用 `/ac on`、`/ac off`、`/ac status`
 
 一句話：
 - `README.md` 是給人 / operator 看的
 - `openclaw-auto-continue/SKILL.md` 是給 agent 看的
+- `references/runtime-integration.md` 是給要落地 runtime 的人 / agent 看的
 
 ---
 
@@ -155,13 +170,28 @@ It also does **not** bypass:
 - destructive action confirmation
 - user decision points
 
+### If another machine says `/ac` is unknown
+That usually means the target machine's **live OpenClaw runtime does not actually implement the command yet**.
+
+So this repo should teach not only usage, but also how to integrate `/ac` into the runtime.
+See:
+
+- `openclaw-auto-continue/references/runtime-integration.md`
+
+That file explains:
+- what the runtime handler must do
+- what `/ac status`, `/ac on`, `/ac off` should change
+- how to verify a real target machine
+- how to troubleshoot a machine where the skill exists but the command does not
+
 ### For agents
 If an OpenClaw agent is deciding how to use this repo:
 
 1. Read `openclaw-auto-continue/SKILL.md`
-2. Follow the trigger conditions and command usage there
+2. If the target machine does not recognize `/ac`, read `openclaw-auto-continue/references/runtime-integration.md`
 3. Use `/ac on`, `/ac off`, and `/ac status` in the relevant chat/session
 
 In short:
 - `README.md` is the human/operator guide
 - `openclaw-auto-continue/SKILL.md` is the agent-facing guide
+- `references/runtime-integration.md` is for runtime integration work

openclaw-auto-continue/SKILL.md

@@ -124,6 +124,8 @@ A practical verification flow:
 3. Give the agent a multi-step but low-risk task
 4. Confirm the agent continues without waiting after every intermediate step
 
+If the target machine does not recognize `/ac`, read `references/runtime-integration.md` and treat it as a runtime integration problem, not a skill-discovery problem.
+
 If you need concrete implementation evidence, the verified live installation path was:
 - `/home/chchang/.nvm/versions/node/v22.22.0/lib/node_modules/openclaw/dist/commands-handlers.runtime-Akj_Dqoi.js`
 

openclaw-auto-continue/references/runtime-integration.md

@@ -0,0 +1,95 @@
+# Runtime Integration: how to actually add `/ac` support
+
+This document is for maintainers/operators who want an OpenClaw installation to truly support `/auto-continue` (`/ac`) at runtime.
+
+## Key point
+
+Installing this skill does **not** automatically add the `/ac` command to a target OpenClaw runtime.
+
+To make `/ac` work on another machine, the running OpenClaw installation must include:
+
+1. command recognition for `/auto-continue` and `/ac`
+2. a command handler that updates the current session's queue settings
+3. persistence of those session settings
+4. queue clearing on disable
+
+## Verified live behavior
+
+The verified implementation used these semantics:
+
+### `/ac status`
+Return the current session queue mode.
+
+### `/ac on`
+Set the current session to:
+- `queueMode = "followup"`
+- `queueDebounceMs = 0`
+- `queueCap = 1`
+- `queueDrop = "summarize"`
+
+### `/ac off`
+- remove those queue settings from the current session
+- clear queued followup/lane items for that session
+
+## Implementation shape
+
+In the verified installation, the runtime had a command handler equivalent to:
+
+- detect `/auto-continue`, `/auto-continue ...`, `/ac`, `/ac ...`
+- authorize sender
+- parse one of: `status | on | off`
+- update `sessionEntry`
+- persist the updated session entry
+- clear queues when switching off
+
+## Verified implementation evidence
+
+A verified live installation contained the handler in:
+
+- `/home/chchang/.nvm/versions/node/v22.22.0/lib/node_modules/openclaw/dist/commands-handlers.runtime-Akj_Dqoi.js`
+
+Observed behavior:
+- `/ac status` returns current queue mode
+- `/ac on` sets followup config
+- `/ac off` clears config and queued items
+
+## What to check on another machine
+
+### 1. Does the running runtime contain the command?
+Do **not** assume a checked-out repo is the active runtime.
+Check the actual installation that the service is using.
+
+### 2. Does the chat surface route commands?
+A skill can exist and still fail if the current surface/runtime path does not process slash commands.
+
+### 3. Does the sender have command permission?
+If command authorization fails, `/ac` may appear to do nothing.
+
+## Minimal verification flow on target machine
+
+1. Install/deploy the runtime that includes `/ac`
+2. In the target chat/session, run:
+   - `/ac status`
+3. If recognized, run:
+   - `/ac on`
+4. Run:
+   - `/ac status`
+5. Give a low-risk multi-step task and verify reduced pausing between steps
+
+## Troubleshooting
+
+### `/ac` is unknown
+Most likely causes:
+- target machine runtime does not include the command yet
+- wrong installation is running
+- command path is unavailable on that surface
+
+### `/ac` exists but does not change behavior
+Check:
+- whether the session setting is actually persisted
+- whether the session is the same one you enabled
+- whether the next step is being blocked by approvals/safety
+
+## One-line deploy rule
+
+**This repo teaches both usage and integration, but `/ac` only works when the target machine's live OpenClaw runtime actually implements the command.**