← Back to Skills Marketplace
192
Downloads
0
Stars
0
Active Installs
8
Versions
Install in OpenClaw
/install verified-task
Description
Enforce correctness before execution. Verify any task output and only proceed if it passes—override requires explicit operator approval.
Usage Guidance
This skill appears coherent and low-risk: it provides procedural guidance to verify outputs and requires explicit human override for non-PASS results. Before installing, ensure your agent actually enforces the human-override requirement and logs overrides (who approved, reason, timestamp). If you integrate any external verification service, avoid sending secrets or sensitive payloads and review that integration separately. The included activator.sh is a harmless reminder script; it does not add hidden behavior.
Capability Analysis
Type: OpenClaw Skill
Name: verified-task
Version: 1.0.7
The 'verified-task' skill is a safety-oriented bundle designed to implement verification guardrails in autonomous agent workflows. It provides instructions (SKILL.md) and templates (DECISION-TEMPLATE.md) to ensure agent outputs are validated against specifications before proceeding, explicitly advising against the handling of secrets or sensitive data. The only executable component, scripts/activator.sh, is a harmless script that prints a text reminder to the console.
Capability Assessment
Purpose & Capability
Name/description (verification guardrail) match the contents: SKILL.md, README, examples, and a small reminder script all describe the same verification gate. No unrelated credentials, binaries, or services are requested.
Instruction Scope
Runtime instructions are limited to defining specs, checking outputs, returning PASS/FAIL/INDETERMINATE, and requiring human overrides. They do not instruct reading system secrets, calling external endpoints, or broad data collection. The optional external verification step is explicitly optional and warns against sending secrets.
Install Mechanism
No install spec; only an instruction-only skill plus a tiny activator.sh reminder (283 bytes). Nothing is downloaded or written to disk by an installer.
Credentials
The skill declares no environment variables, credentials, or config paths. SKILL.md does not reference hidden env vars or other credentials.
Persistence & Privilege
Flags are default (always:false, user-invocable:true, model invocation allowed). The skill does not request permanent/system-wide presence or modify other skills' configs.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install verified-task - After installation, invoke the skill by name or use
/verified-task - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.7
- Added a concrete "Example Usage" section, showing sample input and verification result in JSON format.
- Reformatted lists and sections for improved readability and consistency.
- No changes to core logic or features.
v1.0.6
* Repositioned Verified Task as a **guardrail layer** for agent workflows rather than a generic verification wrapper.
* Introduced **default hard gate behavior**: execution proceeds only on PASS.
* Added **operator-only override rule**; agents cannot override verification results.
* Rewrote `SKILL.md` for clarity, stronger enforcement language, and faster user comprehension.
* Added real-world **use case examples at the top** (payments, content, automation, decisions) to improve adoption.
* Clarified deterministic verification requirements (spec vs output comparison).
* Kept SettlementWitness as an **optional verification layer**, not a dependency.
* Introduced lightweight package structure (“Lightweight+”):
* `README.md` for positioning and usage clarity
* `assets/DECISION-TEMPLATE.md` for consistent PASS/FAIL evaluation
* `assets/USE-CASE-EXAMPLES.md` for practical scenarios
* `references/openclaw-integration.md` for workflow integration guidance
* `scripts/activator.sh` for runtime guardrail reminders
* `_meta.json` for package metadata
* Maintained scan-safe design:
* no required identity
* no credentials or secrets
* no explicit external API calls
* Improved usability while keeping the skill lightweight and easy to integrate across workflows.
v1.0.5
Simplified skill to improve safety and compatibility.
- Removed explicit external endpoint references to reduce scan friction
- Streamlined verification flow for clarity and ease of use
- Clarified decision rules and data handling guidance
- Focused on safe, structured validation pattern for agent workflows
v1.0.4
Strengthened security model: added independent verification guidance, clarified trust requirements, and improved safe data handling. Optimized summary for clarity and adoption.
v1.0.3
Improved summary and top-level description for clarity and higher conversion. No functional changes.
v1.0.2
- Added clear data handling rules: do not send secrets or personal data; only minimal, structured task data required for verification.
- Explicit user approval is now required before submitting output for external verification.
- Clarified privacy/safety: no authentication, secrets, or persistent storage involved; the verifier is public and stateless.
- Updated workflow step: output is submitted for verification only with approval to prevent accidental data exposure.
- Minor clarifications to verification model and process for improved transparency.
v1.0.1
Refined Verified Task for clarity, security, and deterministic verification.
- Added explicit SettlementWitness API endpoint
- Defined verification model and SAR receipt structure
- Clarified that all signing is performed server-side (no keys required)
- Added privacy and safety guarantees (no credentials, no sensitive data)
- Improved instruction clarity and execution flow
- Optimized for security scan confidence and safe agent usage
v1.0.0
Initial release of Verified Task — universal SAR verification pattern for agent workflows.
Metadata
Frequently Asked Questions
What is Verified Task?
Enforce correctness before execution. Verify any task output and only proceed if it passes—override requires explicit operator approval. It is an AI Agent Skill for Claude Code / OpenClaw, with 192 downloads so far.
How do I install Verified Task?
Run "/install verified-task" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Verified Task free?
Yes, Verified Task is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Verified Task support?
Verified Task is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Verified Task?
It is built and maintained by nutstrut (@nutstrut); the current version is v1.0.7.
More Skills