Back to skill

Security audit

custom-llm-provider-setup

Security checks for vulnerabilities and agentic risk

Overview

This skill is a documentation-only guide for configuring Hermes to use user-chosen OpenAI-compatible LLM providers, with expected network and credential handling for that purpose.

Before installing, understand that using this skill can reconfigure Hermes so your prompts, model traffic, and API costs go to a chosen third-party or self-hosted provider. Only enter API tokens for providers you trust, review the final Hermes config after changes, and keep backups when migrating profiles or auxiliary tasks.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (11)

External Script Fetching

High
Category
Supply Chain
Content
format:

```bash
curl -s --request POST \
  --url "https://your-endpoint/v1/chat/completions" \
  --header "Authorization: Bearer $TOKEN" \
  --header "Content-Type: application/json" \
Confidence
90% confidence
Finding
Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

External Script Fetching

High
Category
Supply Chain
Content
Hit the provider's OpenAI-compatible model list endpoint:

```bash
curl -s https://opencode.ai/zen/v1/models | python3 -m json.tool
```

This returns every model the provider serves. The alias will be in the
Confidence
90% confidence
Finding
Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

External Transmission

Medium
Category
Data Exfiltration
Content
format:

```bash
curl -s --request POST \
  --url "https://your-endpoint/v1/chat/completions" \
  --header "Authorization: Bearer $TOKEN" \
  --header "Content-Type: application/json" \
Confidence
60% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
inference. It exposes an **OpenAI-compatible** endpoint at:

```
https://api.cloudflare.com/client/v4/accounts/{ACCOUNT_ID}/ai/v1
```

This means Hermes can use it via `provider: custom` with no plugin.
Confidence
50% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
inference. It exposes an **OpenAI-compatible** endpoint at:

```
https://api.cloudflare.com/client/v4/accounts/{ACCOUNT_ID}/ai/v1
```

This means Hermes can use it via `provider: custom` with no plugin.
Confidence
50% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
inference. It exposes an **OpenAI-compatible** endpoint at:

```
https://api.cloudflare.com/client/v4/accounts/{ACCOUNT_ID}/ai/v1
```

This means Hermes can use it via `provider: custom` with no plugin.
Confidence
50% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
inference. It exposes an **OpenAI-compatible** endpoint at:

```
https://api.cloudflare.com/client/v4/accounts/{ACCOUNT_ID}/ai/v1
```

This means Hermes can use it via `provider: custom` with no plugin.
Confidence
50% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
inference. It exposes an **OpenAI-compatible** endpoint at:

```
https://api.cloudflare.com/client/v4/accounts/{ACCOUNT_ID}/ai/v1
```

This means Hermes can use it via `provider: custom` with no plugin.
Confidence
50% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
`@cf/...` prefixed IDs:

```bash
curl -s --request POST \
  --url "https://api.cloudflare.com/client/v4/accounts/${ACCOUNT_ID}/ai/v1/chat/completions" \
  --header "Authorization: Bearer ${API_TOKEN}" \
  --header "Content-Type: application/json" \
Confidence
60% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
```bash
# Find the default branch
curl -s https://api.github.com/repos/anomalyco/opencode \
  | python3 -c "import json,sys; print(json.load(sys.stdin).get('default_branch','unknown'))"
# → "dev" (not "main")
```
Confidence
50% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Script Fetching

Low
Category
Supply Chain
Content
```bash
# Find the default branch
curl -s https://api.github.com/repos/anomalyco/opencode \
  | python3 -c "import json,sys; print(json.load(sys.stdin).get('default_branch','unknown'))"
# → "dev" (not "main")
```
Confidence
15% confidence
Finding
Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Static analysis

No suspicious patterns detected.