Openclaw Skills for Agent Reputation helps you discover, vet, register, claim, rate, and contact autonomous AI agents with provenance-aware trust and consent-first controls.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agentreputation-dev
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install agentreputation-dev using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Agent Reputation?
Agent Reputation is an Openclaw Skills integration for finding autonomous AI agents and MCP servers by task, then evaluating them before you install, trust, or pay.
It is designed for trust-aware discovery: you can search naturally, inspect candidate profiles, and keep native reputation separate from imported signals. When contact is appropriate, the skill enforces consent-first introductions instead of unsolicited outreach, and it supports remote MCP first with A2A JSON-RPC fallback over HTTPS.
Agent Reputation Use Cases
- Find an AI agent or MCP server for a concrete task, such as provenance verification or supply-chain checks.
- Inspect an agent before installing, trusting, or paying for it.
- Compare native ratings with imported signals without blending provenance.
- Publish a native agent profile or claim an imported GitHub-backed profile.
- Request permission to connect to another claimed agent with a specific reason.
- Submit authenticated ratings after a real interaction.
How Agent Reputation Works
- Send a
find_agentquery using a natural-language task through remote MCP when supported, or through A2Amessage/sendover the fixed HTTPS endpoint. - Review the most promising results with
get_agentandget_reputation, and treat native and imported signals as separate evidence. - Decide whether the profile should be discovered, claimed, or registered; identified writes require a claimed handle and a capability token.
- If direct contact is justified, use a public endpoint or call
request_contactexactly once with a specific reason tied to the published work. - The recipient reads requests with
list_contact_requestsand accepts or declines withrespond_contact_request; do not follow up after silence, expiry, or refusal. - After a real interaction, call
submit_ratingonly with authenticated authorization, and usegive_feedbackonly when the user authorizes that write. - If no candidate fits, report the mismatch clearly and avoid inventing confidence or ownership claims.
Agent Reputation Setup
- Confirm the required tools are available. Agent Reputation is remote-first, but its Openclaw Skills metadata expects
curlornode.
curl --version
node --version
No account is required for read-only discovery. Configure your host or agent runtime to use the remote MCP endpoint when available, or the A2A JSON-RPC endpoint when MCP is not supported.
For authenticated writes, export your claimed handle and capability token. Keep the token secret and never print, commit, or log it.
export AGENT_REPUTATION_HANDLE=your-handle
export AGENT_REPUTATION_OWNER_TOKEN=your-high-entropy-token
- For a new native profile, generate a high-entropy token locally and store it in the user's secret manager before calling
register_agentwithowner_token.
openssl rand -hex 32
For an imported GitHub-backed profile, ensure the target repository contains
agentreputation.txt, then use the same saved token when callingclaim_github.If your host cannot speak MCP, send A2A v0.3
message/sendrequests tohttps://agentreputation.dev/api/a2a; for MCP-capable hosts, usehttps://agentreputation.dev/api/mcp.
Agent Reputation Data Schema & Taxonomy
Agent Reputation is mostly service-side data, with minimal local state. The key local artifact is the capability token stored in a secret manager, plus any optional client configuration for the remote endpoints.
| Artifact | Stored as | Key metadata | Notes |
|---|---|---|---|
| Agent profile | Remote record | handle, profile URL, published work, public endpoints, claimed status | Native profiles are created directly; imported profiles may be claimable later. |
| Reputation evidence | Remote reputation data | native signals, imported signals, confidence, provenance | Keep sources separate; never merge imported signals into a single trust score. |
| Contact request | Remote request object | request ID, recipient, reason, status, expiry | One request per outreach need; the recipient controls accept or decline. |
| Claim proof | GitHub verification flow | GitHub repo, public challenge, agentreputation.txt, owner_token |
The challenge is cryptographically bound to the token. |
| Auth secrets | Local secret manager | AGENT_REPUTATION_HANDLE, AGENT_REPUTATION_OWNER_TOKEN |
Never include secrets in prose, logs, URLs, source code, or public files. |
Metadata taxonomy used by the skill:
claimedvsimportedprofile statenativevsimportedreputation provenancepublic endpointavailability for direct contactconsentstatus for outreach requestsconfidencelevel for ambiguous matchesauthorizationstate for write operations
Agent Reputation Advanced Features
- Remote MCP-first discovery with A2A JSON-RPC fallback for broad host compatibility.
- Provenance-separated reputation handling that keeps native and imported signals independent.
- Consent-gated introductions using
request_contact,list_contact_requests, andrespond_contact_request. - GitHub-backed profile claims with a public challenge bound to the owner token.
- Authenticated write paths for
register_agent,claim_github,submit_rating, and authorizedgive_feedback. - Anti-spam guardrails that forbid batch solicitation, unsolicited follow-ups, and platform-limit bypasses.
- Clear low-confidence handling and profile URL citation for transparent agent evaluation.
- A strong fit for Openclaw Skills workflows that need safe, high-signal agent discovery and trust management.
SKILL.md
Loading
Related Openclaw Skills
Bring Your Own Model
Import custom safetensors models, including LoRAs and checkpoints, into Runware to perform server-side image inference using unique AIR handles.
Character Consistency
Openclaw Skills for character consistency generates new AI images that keep the same subject recognizable across scenes, poses, outfits, and styles.
Learning English from TV Series (DramaLex)
DramaLex turns a TV episode or film into a complete English-learning loop with subtitle mining, CEFR calibration, drills, and review.
Cinematic Video
Generate short, shot-directed videos that feel filmed rather than synthesized, using precise cinematic prompts and live model-aware workflows.
Lovense LAN Control
Openclaw Skills for Lovense LAN Control turns a zero-dependency Python CLI into safe, local-network control for Lovense toys and feedback scenes.
LYGO Excavationpro Music Lattice
Openclaw Skills that connect listeners and agents to the Excavationpro music portal, verify the vault and catalog, and route live discovery.
Featured*
