add-openclaw-agent

Create a new OpenClaw agent via exec (openclaw agents add) and configure its identity + operating rules by editing the new workspace files (IDENTITY.md, AGEN...

Security Scan

The skill's instructions match its stated purpose (creating/configuring an OpenClaw agent) but the package metadata omits important operational requirements (it calls the 'openclaw' CLI via exec and modifies files under ~/.openclaw) and the skill has no source/homepage, so there are unexplained gaps you should understand before installing.

Purpose & Capability

SKILL.md clearly requires running the 'openclaw' CLI (openclaw agents list/add/set-identity) and editing files in ~/.openclaw/workspace-<agent-id>. The registry metadata declares no required binaries/config — that omission is an incoherence (the skill implicitly needs the openclaw CLI and write access to the user's ~/.openclaw workspace).

Instruction Scope

Instructions are narrowly focused: prompt user for purpose, run non-interactive openclaw CLI (after explicit user confirmation), and edit two workspace files (IDENTITY.md and AGENTS.md) with strict append rules. The skill does not instruct reading unrelated system files or exfiltrating data, but it does modify files in the user's home config directory.

Install Mechanism

This is an instruction-only skill (no install spec), which is low-risk in itself. However, because it relies on an external CLI ('openclaw') that is not declared, the runtime must already have that binary available; the metadata should have listed that dependency.

Credentials

The skill does not request any environment variables or credentials in metadata, which is consistent with the described workflow. However, the CLI supports a '--bind <channel[:accountId]>' option (mentioned in SKILL.md) that can route channels/accounts — if used that could require additional account-level permissions not declared here.

Persistence & Privilege

always is false and model invocation is not disabled (platform defaults). The skill writes only to its new agent's workspace files under ~/.openclaw; it does not request persistent platform-wide privileges or modify other skills' configs. Be aware that exec usage can run arbitrary CLI commands if the agent is allowed to act autonomously—this SKILL.md says to seek confirmation before running commands.

What to consider before installing

This skill does what it claims (creates and configures an OpenClaw agent) but has two gaps you should consider before installing: 1) Missing dependency declaration: SKILL.md uses the 'openclaw' CLI via exec, but the package metadata does not list that binary or an install step. Make sure the 'openclaw' CLI is the expected, trusted binary on your system before allowing the skill to run. Ask the publisher to declare this dependency or provide an install spec. 2) File and account effects: the skill will edit files under ~/.openclaw/workspace-<agent-id> and may call CLI flags (e.g., --bind) that route channels/accounts. Review the exact commands the agent will run and confirm each execution when prompted. If you don't trust the skill's source (no homepage/source provided), avoid granting it the ability to run exec or modify your OpenClaw workspace. If the publisher provides a known source/repo or updates metadata to declare the openclaw binary dependency and clarify any account-binding behavior, that would reduce the concern. Otherwise treat this as potentially risky and require explicit confirmation before allowing it to run CLI commands or edit your workspace.

Create a new isolated OpenClaw agent and configure it (identity + role + task) in its workspace.

When this skill applies

• User asks to create a new agent (by name).
• User wants to give that agent an identity (display name, emoji, theme, avatar) or a task/purpose (persona, rules, scope).

Workflow (recommended)

0. Ask what the new agent is for (required)

• Ask the user what they want the new agent to do.
• Ask for any constraints that affect identity and rules (language, tone, safety boundaries, sources allowed, output format, etc.).

Then pick an agent-name based on the purpose. Prefer short, stable ids like:

• beauty, coding, support, ops, research, personal

Notes:

• OpenClaw normalizes names to an agent id (lowercase; invalid characters collapse to -). The id cannot be main (reserved).
• Use the normalized id in the workspace directory: ~/.openclaw/workspace-<agent-id>.

1. Create the agent with exec (non-interactive, JSON)

Use the exec tool to run the OpenClaw CLI. Non-interactive creation requires --workspace and --non-interactive.

Before creating, check whether it already exists:

openclaw agents list --json

Before running creation, explicitly ask for confirmation, for example:

• I am ready to run: openclaw agents add <agent-name> --workspace ~/.openclaw/workspace-<agent-id> --non-interactive --json. Run it now?

Only execute the command after the user confirms.

Create the agent:

openclaw agents add <agent-name> --workspace ~/.openclaw/workspace-<agent-id> --non-interactive --json

Optional flags you can add when needed:

• --model <id> – default model for this agent.
• --agent-dir <dir> – agent state dir (default: ~/.openclaw/agents/<id>/agent).
• --bind <channel[:accountId]> – route a channel to this agent (repeatable).

2. Configure identity by editing IDENTITY.md (then sync)

Edit the new workspace file:

• ~/.openclaw/workspace-<agent-id>/IDENTITY.md

The top section is a template with identity key-value lines (examples):

• name: My Agent
• emoji: 🦞
• theme: short theme
• avatar: path/or/url

Supported keys: name, emoji, theme, creature, vibe, avatar.

Hard rule (IDENTITY.md):

• Only edit the identity field lines (the - **Name:** ... style lines / key-value fields).
• Do not rewrite, delete, or reorder any other template text above the separator.
• Any additional identity/role description must be appended after the --- separator.
• If the file does not contain ---, add a blank line and then a --- line at the end, and append your description after it.
• Tool priority for file changes: prefer edit for targeted append/update; use write only when edit cannot be applied reliably (for example file missing or malformed beyond safe patching).

Short example (append style):

... (keep/update existing identity fields)

---

我负责：根据肤质/预算/场景给出 2-3 个可选方案，并说明优缺点与使用顺序。

After editing, sync to config via exec:

openclaw agents set-identity --workspace ~/.openclaw/workspace-<agent-id> --from-identity --json

3. Configure role + task + operating rules in AGENTS.md

Edit the new workspace file:

• ~/.openclaw/workspace-<agent-id>/AGENTS.md

Put task definition and operating rules here (how to work), based on what the user told you in step 0. Examples of good content:

• What this agent is responsible for (scope)
• What it must not do (red lines)
• The question-asking checklist before acting
• Output format expectations (bullets, structure, brevity)
• Tooling rules (when to use exec; avoid destructive commands unless asked)

Hard rule (AGENTS.md):

• Do not rewrite, delete, or reorder the default template content.
• Add all custom “role/task/rules” content only by appending after the --- separator.
• If the file has no ---, add a blank line + --- near the end, then append your custom section after it.
• Tool priority for file changes: prefer edit for targeted append/update; use write only when edit cannot be applied reliably (for example file missing or malformed beyond safe patching).

Short example (append style):

... (keep default template unchanged)

---

## 美妆助手工作规范
- 先问：肤质/预算/诉求/过敏史
- 再给：2-3 个方案 + 优缺点 + 价格区间
- 禁止：假装亲测；推荐三无/杂牌

If you also want a stable persona/tone/boundary file, place that in SOUL.md too, but keep the core “how to operate” rules in AGENTS.md.

End-to-end flow (summary)

1. Ask what the new agent is for (purpose + constraints).
2. Choose <agent-name> (short, stable; will normalize to <agent-id>).
3. openclaw agents add <agent-name> --workspace ~/.openclaw/workspace-<agent-id> --non-interactive --json
4. Edit ~/.openclaw/workspace-<agent-id>/IDENTITY.md
5. openclaw agents set-identity --workspace ~/.openclaw/workspace-<agent-id> --from-identity --json
6. Edit ~/.openclaw/workspace-<agent-id>/AGENTS.md with the role/task/rules.

References

• Multi-agent routing and paths: Multi-Agent Routing, Agent workspace.
• CLI: openclaw agents add, openclaw agents set-identity, openclaw agents bind in agents CLI.