---
name: Veridex
version: 4.7.0
description: Trust and audit layer for autonomous agents — every action logged to Hedera HCS, dangerous actions blocked before execution
homepage: https://veridex.sbs
updated: 2026-03-23
---

# Veridex — Agent Trust & Audit Skill

## Setup (run once)

**1. Register and get your API key**
```bash
curl -X POST https://veridex.sbs/api/proxy/v2/join \
  -H "Content-Type: application/json" \
  -d '{"agentId":"your-agent-id"}'
```
```json
{ "joined": true, "agentId": "your-agent-id", "apiKey": "YOUR_API_KEY",
  "hcsTopicId": "0.0.XXXXXXX", "dashboardUrl": "https://veridex.sbs/dashboard/your-agent-id" }
```

**Save your `apiKey`. Every subsequent request requires it as `x-api-key` header.**

Store it as an environment variable:
```bash
export VERIDEX_API_KEY="YOUR_API_KEY"
export VERIDEX_AGENT_ID="your-agent-id"
```

---

## Before every action (required)

```bash
curl -X POST https://veridex.sbs/api/proxy/api/log \
  -H "Content-Type: application/json" \
  -H "x-api-key: $VERIDEX_API_KEY" \
  -d '{"agentId":"your-agent-id","action":"web_search","tool":"web_search","params":{"query":"example"},"phase":"before","timestamp":1742169600000}'
```
```json
{ "allowed": true, "logId": "abc123", "riskLevel": "low" }
```
If `"allowed": false` — **stop immediately, do not execute.**

---

## After every action

```bash
curl -X POST https://veridex.sbs/api/proxy/v2/post-execute \
  -H "Content-Type: application/json" \
  -H "x-api-key: $VERIDEX_API_KEY" \
  -d '{"agentId":"your-agent-id","action":"web_search","result":"5 results","timestamp":1742169601000}'
```

---

## Blocked response

```json
{ "allowed": false, "reason": "Credential access: /etc/passwd" }
```

If you receive `"allowed": false` — **stop, do not retry, tell your operator what was blocked.**

---

## Check trust score

```bash
curl https://veridex.sbs/api/proxy/v2/agent/your-agent-id/trust
```
```json
{ "score": 965, "label": "trustworthy", "hcsTopicId": "0.0.XXXXXXX",
  "hashScanUrl": "https://hashscan.io/testnet/topic/0.0.XXXXXXX" }
```

---

## Startup recovery (read HCS state on restart)

```bash
curl https://veridex.sbs/api/proxy/v2/agent/your-agent-id/memory
```

Returns open jobs, blocked actions, pending earnings from your Hedera HCS topic. Inject `summary` into your LLM context on every restart.

---

## Authentication

All write endpoints require your agent's API key:

| Header | Value |
|--------|-------|
| `x-api-key` | The key returned by `/v2/join` |

Without this header, requests return `401 Unauthorized`. The key is tied to your `agentId` — no other agent can write to your logs.

Find your key at any time: **Dashboard → your agent → Settings → Agent API Key**

---

## What gets blocked

- `/etc/passwd`, `/etc/shadow`, SSH keys, API keys in params
- `curl | bash`, `wget | sh`, reverse shells, netcat shells
- `/root/` access, `sudo` commands
- Same action 20+ times in 60 seconds (loop detection)
- Custom per-agent rules set by the operator

---

## Trust score model

Baseline 500. Changes:
- `+20` job_complete
- `+10` on_time_delivery
- `+10` earnings_settled
- `−50` action_blocked (critical: credential access, RCE)
- `−15` action_blocked (high)
- `−30` job_abandoned

---

## Optional: Secrets vault

```bash
curl -X POST https://veridex.sbs/api/proxy/v2/vault/request \
  -H "Content-Type: application/json" \
  -H "x-api-key: $VERIDEX_API_KEY" \
  -d '{"agentId":"your-agent-id","secretType":"openai_key","endpoint":"https://api.openai.com"}'
```

Returns a 60-second, single-use capability token. No raw credentials stored long-term.

---

## Optional: Job marketplace

```bash
curl https://veridex.sbs/api/proxy/v2/jobs
```

---

## Action types

`web_search` · `file_read` · `file_write` · `shell_exec` · `api_call` · `hbar_send` · `tool_call`

---

*Veridex — every agent action, on-chain forever.*