More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    🛡️ OpenClaw Guardrail Linter

    2

    Review selected OpenClaw configuration for risky declarations, with explicit reporting when inputs cannot be assessed.

    $12

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    🛡️ OpenClaw Guardrail Linter

    🛡️ OpenClaw Guardrail Linter

    Example session with this skill installed

    Review only my selected stable OpenClaw configuration files. Report risky declarations and anything that could not be assessed. Do not execute OpenClaw or change configuration.

    • Read your context and instructions
    • Compiled the openclaw guardrail linter

    Illustrative v1.1.0 result guide, not a live-host audit

    Use the bundled clean and risky fixtures first. For your selected inputs, review the emitted findings and coverage status together; an incomplete assessment must not be treated as clean.

    ResultMeaningNext step
    Exit 0No findings in assessed inputsReview coverage and the tool limitations
    Exit 1Patterns need reviewInspect the cited rule and relevant declaration
    Exit 2Assessment incompleteResolve the input or coverage problem before drawing conclusions

    Reports identify inputs by ordinal and source hash rather than exposing raw values or absolute paths. Line 0 means source line mapping is unavailable, not that a precise line was verified.

    What this does not establish

    The linter does not resolve inherited policies, follow includes, or prove effective runtime permissions. Confirm actual host behavior separately. Review any proposed remediation and back up customizations before making changes.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Review declarations, not assumed protection

    OpenClaw Guardrail Linter reads selected local configuration and instruction files and flags patterns worth reviewing: broad tool declarations, open channel settings, literal credential-like values, incomplete file-tool denials and sandbox-related settings. It does not run OpenClaw, change configuration or calculate the running agent's effective permissions.

    Current package

    Package v1.1.0 is listed in Agensi revision v1.1. The changes and setup below apply to this version; inspect the version before updating an older installation.

    What changes in v1.1.0

    Parsed JSON5 replaces regex-only configuration interpretation. The update fixes missed file-mutation and neighboring-SecretRef findings and distinguishes incomplete input from a scan with no findings. Linked/reparse paths, missing or malformed inputs, unsupported includes and files above 64 KiB are not reported as clean. Credential/state paths are excluded without reading their contents. Reports withhold raw values and absolute paths, using input ordinals and source hashes. Source line mapping is unavailable (line 0).

    Setup and first use for v1.1.0

    Requires Python 3.10+ and json5==0.15.0. Extract the complete folder and install the pinned dependency in an approved environment with python -m pip install -r requirements.txt. Setup requires package-source access; scans are offline. No API key or live OpenClaw connection is required.

    Run python scripts/scan_openclaw_guardrails.py fixtures/clean-sample, then the risky fixture before selecting your own stable configuration. Exit 0 means no findings in assessed inputs, 1 means findings, and 2 means incomplete assessment. None is a security certification.

    Limits

    Not a complete host schema validator, vulnerability scanner or inherited-policy resolver. Includes are not followed. Other policy layers can constrain broad declarations; allowed shell execution can write files despite file-tool denials. Use native host policy/audit evidence for actual execution. No live-host enforcement or adversarial concurrent-replacement resistance is claimed. Back up customizations and check automation that previously expected only exit 0/1.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Security.

    Static declaration review, not a full host-schema validator, vulnerability scanner or effective-permissions resolver. Includes are not followed. v1.1.0 uses exit 2 for incomplete assessment and line 0 when source mapping is unavailable. No live-host enforcement claim or adversarial file-replacement guarantee.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    Software
    Python 3.10+json5==0.15.0

    Permissions

    Read Files

    File Scopes

    **/openclaw.json
    **/.openclaw/**/*.json
    **/.openclaw/**/*.json5
    **/.config/openclaw/**/*.json
    **/SOUL.md
    **/AGENTS.md
    **/USER.md

    Read Files only. It reads your OpenClaw config and workspace files as plain text to flag unsafe settings and committed secrets. It does not run OpenClaw, read your live environment, use secrets, or send anything anywhere. No write, shell, or network.

    Listed4 months ago
    Updated20 days ago

    What's inside

    Frequently Asked Questions

    The linter identifies 11 specific vulnerability classes (OGL001-OGL011) unique to OpenClaw, such as sandbox escapes, Docker bind mount risks, and the 'deny-write' bypass that standard security tools often miss.

    It supports all standard OpenClaw configuration formats including .json and .json5, and it specifically analyzes specialized workspace manifests like SOUL.md, AGENTS.md, and USER.md.

    The skill performs a static heuristic analysis of your configuration files, meaning it reads your text-based settings without requiring access to your live gateway or network environment.

    Updates are released as the OpenClaw framework evolves; buyers receive lifetime access to all future rule set expansions and new vulnerability definitions for this version.

    By auditing tool-use permissions and DM settings, the linter ensures that agents aren't granted excessive 'exec' power or open inbound communication channels that could lead to unauthorized system access.

    New in Security