Feishu Agent Skills

Create and wire a new OpenClaw agent with a fixed workflow. Use when the user asks to create/add a new OpenClaw agent or says “我要创建一个新的 Agent”, automate mult...

Security Scan

The skill's behavior largely matches its stated purpose (creating/wiring OpenClaw Feishu agents), but there is a discrepancy: the runtime instructions require the 'openclaw' CLI and will read/write ~/.openclaw/openclaw.json (including writing app secrets), yet the skill metadata does not declare the CLI requirement or any credential/environment needs — this mismatch and the sensitive file writes deserve attention before installing.

Purpose & Capability

Name/description, SKILL.md, and included script all target creating/updating OpenClaw agents for the Feishu channel — behavior is coherent. However, the skill's metadata lists no required binaries while the instructions explicitly require the 'openclaw' CLI to be present and used; that's an omission/mismatch in declared requirements.

Instruction Scope

The SKILL.md steps are narrowly scoped to OpenClaw agent creation: reading/backing up ~/.openclaw/openclaw.json, running openclaw CLI commands, following logs to extract peer IDs, and running the bundled Python script to upsert accounts/bindings/dmScope. No external network endpoints or unexpected data exfiltration are present in the instructions.

Install Mechanism

No install spec; this is instruction-only with a small included Python script. There are no downloads from external URLs and no packages installed automatically, so install risk is low.

Credentials

The skill requests sensitive values (appId/appSecret) as command-line arguments and will write them into ~/.openclaw/openclaw.json via the script. The registry declares no required env vars or primary credential, which is consistent with passing secrets as args, but users should be aware that secrets will be stored in the config file (potentially insecure depending on filesystem permissions).

Persistence & Privilege

The skill is not always-enabled and uses normal agent invocation. It does not request to modify other skills or system-wide agent settings beyond editing the user's OpenClaw config file, backing it up, and invoking OpenClaw CLI commands — privileges are limited to the user's filesystem and running CLI tools.

What to consider before installing

Before installing or running: (1) Confirm the 'openclaw' CLI is installed and you intend to use it — the SKILL.md requires it even though metadata doesn't declare it. (2) Review the bundled script (scripts/upsert_openclaw_agent.py) yourself — it reads and overwrites ~/.openclaw/openclaw.json and will store appSecret/appId/botName there. (3) Run the script in dry-run mode first (--dry-run) to see proposed changes, and keep backups (the script can create one). (4) Check file permissions on ~/.openclaw/openclaw.json if you will store secrets there; consider using more secure secret storage if needed. (5) If you don't trust the skill source, perform these steps manually (use openclaw agents add and edit openclaw.json by hand) rather than granting the skill automatic execution.

OpenClaw Create Agent

Use this skill to create one new OpenClaw agent and finish routing/config updates in one pass.

This skill is Feishu-only. Do not use it for Telegram, Slack, Discord, or other channels.

Required Inputs

First align the user's target scenario, then collect fields. Ask only for missing required fields.

Scenario alignment question (must ask first):

• Option A: create a new bot/app and map it to a new agent (routing_mode=account)
• Option B: map an existing/new group to a new agent under one bot (routing_mode=peer)

Field	Required	Notes
agent_id	yes	lowercase letters, digits, - only
workspace	yes	use absolute path for project workspace
model	no	optional model override for this agent
routing_mode	yes	account or peer
channel	no	fixed to feishu

For routing_mode=account (one bot per agent), collect:

• account_id (required)
• app_id, app_secret, bot_name (optional but recommended)

For routing_mode=peer (single bot, multi-group), collect:

• peer_kind (group or direct, required)
• peer_id (required)
• account_id (optional, narrows matching if provided)
• if this is a new group, collect peer_id only after guided gates in Step 0.5

Workflow

1. Align scenario (new bot per agent or multi-group per agent).
2. For multi-group scenarios, run guided gates in Step 0.5.
3. Read current config and backup.
4. Create agent runtime with OpenClaw CLI.
5. Upsert config for channel account, binding, and dmScope.
6. Restart gateway and verify.
7. Return a concise change summary.

Step 0: Align Requirement

Ask this first:

你是要哪一种？
1) 新建一个机器人，对应一个新 Agent（account 路由）
2) 已有一个机器人，在新群里绑定一个新 Agent（peer 路由）

Map answer:

• option 1 -> routing_mode=account
• option 2 -> routing_mode=peer

If answer is ambiguous, stop and clarify before editing config.

Step 0.5: Multi-Group Guided Gates

Apply this step when the user is doing multi-group multi-agent routing.

1. Guide the user to create a new group and add the bot into that group.
2. Pause and wait for explicit confirmation: 已创建群.
3. Only after receiving 已创建群, guide the user to send one message in that group to generate logs.
4. Pause and wait for explicit confirmation: 已发送.
5. Only after receiving 已发送, check OpenClaw logs and extract chat_id (format oc_xxxxx).

Suggested command:

openclaw logs --follow

Expected signal in logs:

Received message from peer: { kind: "group", id: "oc_xxxxxxxxxxxxxxxx" }

Step 1: Precheck and Backup

Run:

test -f ~/.openclaw/openclaw.json
cp ~/.openclaw/openclaw.json ~/.openclaw/openclaw.json.bak.$(date +%Y%m%d%H%M%S)

If config file is missing, stop and ask user to initialize OpenClaw first.

Step 2: Create Agent Runtime

Run OpenClaw CLI first (required):

openclaw agents add <agent_id> --workspace <workspace> --non-interactive

If model is provided:

openclaw agents add <agent_id> --workspace <workspace> --model <model> --non-interactive

Step 3: Upsert openclaw.json

Use the bundled script:

python3 scripts/upsert_openclaw_agent.py \
  --config ~/.openclaw/openclaw.json \
  --agent-id <agent_id> \
  --routing-mode <account|peer> \
  [--account-id <account_id>] \
  [--app-id <app_id>] \
  [--app-secret <app_secret>] \
  [--bot-name <bot_name>] \
  [--peer-kind <group|direct>] \
  [--peer-id <peer_id>]

Run from this skill folder, or replace scripts/upsert_openclaw_agent.py with its absolute path.

Script behavior:

• upsert channels.<channel>.accounts.<account_id> when account mode is used
• upsert one binding for this agent
• always enforce session.dmScope = per-account-channel-peer for multi-agent Feishu setup

Read references/routing-modes.md when routing choice is unclear.

Step 4: Restart and Verify

Run:

openclaw gateway restart
openclaw agents list --bindings

Validate:

• target agent appears in list
• binding points to expected channel/account or channel/peer

Step 5: Report Output

Return:

• created/updated binding match
• whether account entry was added/updated
• whether dmScope changed
• verification command results

Constraints

• Preserve existing unrelated agents/accounts/bindings.
• Reject route conflicts (same channel/account/peer route already used by another agent).
• Avoid interactive prompts unless user explicitly asks for interactive mode.
• Assume channel=feishu only.
• In multi-group scenarios, do not continue past Step 0.5 until receiving 已创建群 and then 已发送.