Ide-IM-Skill

You are managing the Ide-IM-Skill bridge (IM → AI 桥接，可在 Cursor 内启动与管理). User data path is controlled by CTI_HOME (default ~/.ide-im/). 配哪个就用哪个路径。

The skill directory (SKILL_DIR) is the directory containing this SKILL.md. Resolve it by: Glob for **/Ide-IM-Skill/SKILL.md or **/ide-im-skill/SKILL.md and use the parent of the file's directory.

Command parsing

Parse the user's intent into one of these subcommands:

Disambiguation: status vs doctor — Use status for "is it running?". Use doctor when the user reports a problem.

Extract optional numeric argument for logs (default 50).

Before asking for platform credentials, read SKILL_DIR/references/setup-guides.md and present the relevant guide to the user.

Installation

Install via skills.sh (no git clone needed):

npx skills add chovizzz/Ide-IM-Skill

Dependencies are auto-installed on first start — no manual npm install required.

Runtime detection (Cursor vs others)

• Cursor — You are in Cursor. AskUserQuestion may be available for interactive setup. SKILL_DIR is the folder containing this SKILL.md (e.g. Ide-IM-Skill or ide-im-skill).
• No AskUserQuestion — Show SKILL_DIR/config.env.example and instruct the user to create $CTI_HOME/config.env (default ~/.ide-im/config.env) manually.

Config check (for start, stop, status, logs, reconfigure, doctor)

If config.env does NOT exist (under CTI_HOME, default ~/.ide-im):

• With AskUserQuestion: start the setup wizard.
• Without: show SKILL_DIR/config.env.example and ask the user to create the file; do not start the daemon.

If it exists, proceed with the requested subcommand.

Subcommands

setup

Run an interactive setup wizard when AskUserQuestion is available. Otherwise show SKILL_DIR/config.env.example with field-by-field explanations.

When interactive, collect one field at a time, confirm each value (mask secrets to last 4 chars).

Step 1 — Choose channels

Ask which channels to enable: telegram, discord, feishu, qq (comma-separated). Briefly:

• telegram — Personal use, streaming, inline permission buttons.
• discord — Team use, server/channel access control.
• feishu (Lark) — Feishu/Lark teams, event-based.
• qq — QQ C2C only, text /perm for permissions.

Step 2 — Collect tokens per channel

For each enabled channel, use SKILL_DIR/references/setup-guides.md. Collect:

• Telegram: Bot Token → Chat ID → Allowed User IDs (optional). At least Chat ID or Allowed Users required.
• Discord: Bot Token → Allowed User IDs → Allowed Channel IDs / Guild IDs (optional). At least one of Allowed Users or Allowed Channels required.
  • Important: In the Discord Developer Portal → Bot → Privileged Gateway Intents, enable MESSAGE CONTENT INTENT. Without it, guild messages have empty content and will be silently dropped.
  • Require Mention (CTI_DISCORD_REQUIRE_MENTION): set true to only respond when the bot is @mentioned in guild channels (default false).
  • Group Policy (CTI_DISCORD_GROUP_POLICY): open (default, respond to guild messages) or disabled (completely ignore guild messages).
• Feishu: App ID → App Secret → Domain (optional) → Allowed User IDs (optional). Guide through permissions, bot, events (long connection), publish.
  • Require Mention (CTI_FEISHU_REQUIRE_MENTION): default true for Feishu — group messages need @mention; set false to respond to all messages.
  • Group Policy (CTI_FEISHU_GROUP_POLICY): open (default) or disabled (ignore all group messages).
• QQ: App ID → App Secret → Allowed User OpenIDs (optional) → Image Enabled / Max Image Size (optional). Remind: C2C only, no inline buttons.

Step 3 — General settings

• Runtime: claude, codex, auto, or cursor
  • claude — Claude Code CLI + Claude Agent SDK
  • codex — OpenAI Codex SDK
  • auto — Try Claude first, fall back to Codex
  • cursor — Use Cursor CLI (agent); install: curl https://cursor.com/install -fsS | bash

Step 3a — CLI check & auto-install (immediately after runtime selection)

After the user picks a runtime, check whether the required CLI exists and auto-install if missing.

Flow (use SKILL_DIR/scripts/check-cli.sh):

1. Run: bash "SKILL_DIR/scripts/check-cli.sh" <runtime> — outputs found: /path (version: x) or not_found.
2. If found → show path and version, proceed.
3. If NOT found → tell the user the CLI is missing and ask "是否自动安装？(Y/n)".
   • If yes → run: bash "SKILL_DIR/scripts/check-cli.sh" <runtime> --install — attempts install then re-checks.
   • If no → show manual install instructions and continue (the daemon will fail at start if CLI is still missing).
4. For auto: the script checks both claude and codex; at least one must succeed.

• Working Directory:
  • If CTI_DEFAULT_WORKDIR is set in config.env,始终优先使用它。
  • 如果未设置：
    • Runtime = cursor 时，默认 ~/.workspace（每个 Cursor 工程共享的全局工作区目录）。
    • 其它 runtime（claude / codex / auto）保持上游行为：使用当前进程的工作目录 $CWD。
• Identity/Memory 目录（SOUL.md、AGENTS.md、MEMORY.md 等）：
  • 若设置 CTI_IDENTITY_DIR 则使用该路径。
  • Runtime = cursor 且未设置时，默认使用 ~/.workspace；首次启动会从 templates/identity-default/ 自动填充。会话出生时会在 system prompt 里写明「Workspace (identity root): <路径>」，和 OpenClaw 一样让 agent 知道工作区在哪。
  • 其它 runtime 未设置时，使用各会话的 working directory 作为 identity root。
• Model (optional): leave blank to use runtime default
• Mode: code, plan, ask

Step 3b — Workspace & identity initialisation (auto, immediately after CLI check)

After CLI is verified, automatically create workspace directories and seed identity templates.

Run: bash "SKILL_DIR/scripts/init-workspace.sh" <runtime>

This creates:

1. CTI_HOME dirs (~/.ide-im/{data,logs,runtime,data/messages})
2. Working directory (~/.workspace for cursor, $CWD for others)
3. Identity directory (cursor only: ~/.workspace/ + memory/) — seeds OpenClaw templates (AGENTS.md, SOUL.md, IDENTITY.md, USER.md, TOOLS.md, MEMORY.md) from templates/identity-default/ if not already present

If the user set custom CTI_DEFAULT_WORKDIR or CTI_IDENTITY_DIR, pass them:
bash "SKILL_DIR/scripts/init-workspace.sh" <runtime> --work-dir <path> --identity-dir <path>

After init, remind the user they can personalise USER.md and MEMORY.md in the identity directory.

Long session memory + Discord streaming (optional)

• CTI_MEMORY_COMPRESS_AFTER_MESSAGES (default 24): when the bridge has stored this many messages for a channel session, it runs a background summarization into MEMORY.md under the identity dir, then truncates stored messages to CTI_MEMORY_KEEP_AFTER_COMPRESS (default 4). The next user turn still sees long-term facts via the normal MEMORY.md injection in system prompt. Set to 0 to disable.
• CTI_DISCORD_STREAM_ENABLED (default true): Discord streams partial replies while the model runs. Set false if you ever see duplicate final messages.

Step 4 — Write config and validate

1. Summary table (secrets masked).
2. Confirm, then: mkdir -p "$CTI_HOME"/{data,logs,runtime,data/messages} (CTI_HOME 未设时用 ~/.ide-im)
3. Write $CTI_HOME/config.env (KEY=VALUE).
4. chmod 600 $CTI_HOME/config.env
5. Validate tokens per SKILL_DIR/references/token-validation.md
6. On success: "Setup complete! Run ide-im start (or say «启动桥接») to start the bridge."

start

Ensure config.env exists under CTI_HOME. Then:

bash "SKILL_DIR/scripts/daemon.sh" start

On failure, suggest ide-im doctor and ide-im logs.

stop

bash "SKILL_DIR/scripts/daemon.sh" stop

status

bash "SKILL_DIR/scripts/daemon.sh" status

logs

Optional line count N (default 50): bash "SKILL_DIR/scripts/daemon.sh" logs N

reconfigure

1. Read config from CTI_HOME/config.env, show table (secrets masked).
2. Ask what to change, update config atomically (tmp + rename).
3. Re-validate changed tokens.
4. Remind: "Run ide-im stop then ide-im start to apply."

doctor

bash "SKILL_DIR/scripts/doctor.sh"

Suggest fixes; for details see SKILL_DIR/references/troubleshooting.md.

Notes

• Mask secrets (last 4 chars only) in all output.
• Always check config.env exists before starting the daemon.
• Config path: CTI_HOME/config.env（未设 CTI_HOME 时为 ~/.ide-im/config.env）。配哪个就用哪个。