Skills config

Most skills configuration lives under skills in ~/.openclaw/openclaw.json. Agent-specific visibility lives under agents.defaults.skills and agents.entries.*.skills.

{  skills: {    allowBundled: ["gemini", "peekaboo"],    load: {      extraDirs: ["~/path/to/agent-scripts/skills"],      allowSymlinkTargets: ["~/path/to/skills"],      watch: true,    },    install: {      preferBrew: true,      nodeManager: "npm",      allowUploadedArchives: false,    },    workshop: {      autonomous: { mode: "auto" },      approvalPolicy: "auto",      maxPending: 50,      maxSkillBytes: 40000,    },    entries: {      "image-lab": {        enabled: true,        apiKey: { source: "env", provider: "default", id: "GEMINI_API_KEY" },        env: { GEMINI_API_KEY: "GEMINI_KEY_HERE" },      },      peekaboo: { enabled: true },      sag: { enabled: false },    },  },}

Loading (skills.load)

Install (skills.install)

Operator Install Policy (security.installPolicy)

Use security.installPolicy when operators need a trusted local command to approve or block skill and plugin installs with host-specific policy. The policy runs after OpenClaw has staged source material and before the install or update continues. It applies to ClawHub skills, uploaded skills, Git/local skills, skill dependency installers, and plugin install/update sources.

{  security: {    installPolicy: {      enabled: true,      // Omit targets to cover every supported target.      targets: ["skill", "plugin"],      exec: {        source: "exec",        command: "/usr/local/bin/openclaw-install-policy",        args: ["--json"],        timeoutMs: 10000,        noOutputTimeoutMs: 10000,        maxOutputBytes: 1048576,        passEnv: ["OPENCLAW_STATE_DIR", "PATH"],        env: { POLICY_MODE: "strict" },        trustedDirs: ["/usr/local/bin"],      },    },  },}

The policy command and interpreter script arguments must be direct regular files with trusted ownership, restricted permissions, and verifiable parent directories. Symlinks and insecure paths are rejected.

The policy receives one JSON object on stdin with protocolVersion: 1, openclawVersion, targetType, targetName, sourcePath, sourcePathKind, optional structured source, structured origin, and request. It must write one JSON object on stdout with an allow, warn, or block decision. warn and block require a non-empty reason; every decision may include a findings array. Each finding requires non-empty string ruleId and message fields plus a severity of info, warn, or critical. Optional file and evidence values must be non-empty strings; a finite numeric line is rounded down and clamped to the safe-integer range from 1 through Number.MAX_SAFE_INTEGER. Malformed finding entries are ignored, and invalid optional fields are omitted. A non-array findings value is treated as absent. Operator-facing reason and finding text are limited to 1,000 characters. OpenClaw retains at most 100 normalized findings for display. Only a warn response with more than 100 valid findings fails closed and cannot be acknowledged; allow and block retain the first 100. A warning stops the install before commit. A warn review whose fully rendered notice, including its title, target, sanitized reason and findings, and recovery guidance, exceeds the 4,000-character aggregate display limit fails closed without presenting a partial review. An over-budget block remains terminal with a bounded denial, while over-budget findings on allow are summarized in bounded diagnostic output. Interactive CLI plugin and skill commands ask the operator to type the target name using the same install anyway or update anyway copy as suspicious ClawHub releases, then run policy again before continuing. Declined and non-interactive commands on the direct CLI may use --acknowledge-install-policy-warning as explicit approval after review for every warning in that command invocation; every approved warning is re-evaluated before continuing. The Control UI can review and approve warnings for its plugin install request; that approval covers every warning in the invocation, and each warning is still re-evaluated. Other Gateway-backed and automatic installs remain blocked when they have no operator-confirmation flow. Use an equivalent direct plugin or skill command to review and approve the warning when one exists. Otherwise, change security.installPolicy to return allow for the reviewed request, then retry the managed flow. --force does not approve policy warnings. A block, non-zero exit, timeout, invalid JSON, non-object response, missing or invalid protocol version or decision, or missing or empty warn/block reason always fails closed.

OpenClaw does not execute install policy during normal Gateway startup. Installs and updates fail closed when policy is enabled but unavailable. openclaw doctor performs static validation; openclaw doctor --deep executes a synthetic install probe against the configured command.

Bulk updates apply policy per target: a blocked skill or plugin update fails that target without disabling the policy or skipping later targets in the batch.

Example stdin:

{  "protocolVersion": 1,  "openclawVersion": "2026.6.1",  "targetType": "skill",  "targetName": "weather",  "sourcePath": "/var/folders/.../openclaw-skill-clawhub/root",  "sourcePathKind": "directory",  "source": {    "kind": "clawhub",    "authority": "openclaw",    "mutable": false,    "network": true  },  "origin": {    "type": "clawhub",    "registry": "https://clawhub.openclaw.ai",    "slug": "weather",    "version": "1.0.0"  },  "request": {    "kind": "skill-install",    "mode": "install",    "requestedSpecifier": "clawhub:[email protected]"  },  "skill": {    "installId": "clawhub"  }}

Minimal policy command:

#!/usr/bin/env node let input = "";process.stdin.setEncoding("utf8");process.stdin.on("data", (chunk) => {  input += chunk;});process.stdin.on("end", () => {  const request = JSON.parse(input);  if (request.targetType === "plugin" && request.source?.kind === "local-path") {    process.stdout.write(      JSON.stringify({        protocolVersion: 1,        decision: "block",        reason: "local plugin paths are not approved on this host",      }),    );    return;  }  process.stdout.write(JSON.stringify({ protocolVersion: 1, decision: "allow" }));});

Bundled skill allowlist

Per-skill entries (skills.entries)

Keys under entries match the skill name by default. If a skill defines metadata.openclaw.skillKey, use that key instead. Quote hyphenated names (JSON5 allows quoted keys).

Agent allowlists (agents)

Use agent config when you want the same machine/workspace skill roots but a different visible skill set per agent.

{  agents: {    defaults: {      skills: ["github", "weather"], // shared baseline    },    entries: {      writer: { default: true }, // inherits github, weather      docs: { skills: ["docs-search"] }, // replaces defaults entirely      "locked-down": { skills: [] }, // no skills    },  },}

Workshop (skills.workshop)

See Self-learning for eligibility, privacy, cost, proposal-only permissions, and troubleshooting.

See Skill Workshop for the proposal lifecycle, CLI commands, agent tool parameters, and Gateway methods this config controls.

Symlinked skill roots

By default, workspace, project-agent, extra-dir, and bundled skill roots are containment boundaries. A symlinked skill folder under <workspace>/skills that resolves outside the root is skipped with a log message.

To allow an intentional symlink layout, declare the trusted target:

{  skills: {    load: {      extraDirs: ["~/path/to/skills"],      allowSymlinkTargets: ["~/path/to/skills"],    },  },}

With this config, <workspace>/skills/manager -> ~/path/to/skills is accepted after realpath resolution. extraDirs scans the sibling repo directly; allowSymlinkTargets preserves the symlinked path for existing layouts.

Skill Workshop uses each agent's <state-dir>/agents/<agentId>/agent/workshop-skills containment boundary. It does not use allowSymlinkTargets, and it rejects symlinked skills that resolve outside that directory.

Managed ~/.openclaw/skills and personal ~/.agents/skills directories already accept skill-directory symlinks unconditionally (per-skill SKILL.md containment still applies) — allowSymlinkTargets is only needed for workspace, extra-dir, and project-agent (<workspace>/.agents/skills) roots.

Sandboxed skills and env vars

Pass secrets into a Docker sandbox with:

{  agents: {    defaults: {      sandbox: {        docker: {          env: { GEMINI_API_KEY: "your-key-here" },        },      },    },  },}

Loading order reminder

See Loading order for source precedence, including the per-agent Workshop tier, and Snapshots and refresh for when changes become visible.

Related