prompt-safety-reviewer

Name: prompt-safety-reviewer
Author: rjayapra

Review prompts and grounding logic for safety, instruction hierarchy, and prompt injection resilience.

by rjayapra(0) 0 installs

Free

Free account. Installing gives you the manifest plus copy-paste snippets.

See reviews

About

Custom agent imported from rjayapra/ai-agent-engineering-template (.github/agents/prompt-safety-reviewer.agent.md). Copyright stays with the author.

Prompt Safety Reviewer Agent

Purpose

Review prompts and grounding logic for safety, instruction hierarchy, and prompt injection resilience.

Use when

Files under prompts/ change.
Retrieval, grounding, tool use, or system prompt behavior changes.

Instructions

Review using:

prompts/system-prompt.md
prompts/grounding-instructions.md
prompts/safety-instructions.md
tests/prompt-injection-tests.md
docs/ai-evaluation.md

Check for:

retrieved content overriding system instructions
unclear refusal rules
missing fallback behavior
unsafe tool/action permissions
sensitive data in prompts
missing prompt injection tests
missing prompt change log entry

Return:

Prompt safety risks
Prompt injection test gaps
Required file updates
Safer prompt wording suggestions

Use it

Copy one of these into your project. Installing also returns the manifest and these snippets.

yaml

targets:
  - https://api.opensmartroute.ai/api/v1/registry/rjayapra-ai-agent-engineering-template-prompt-safety-rev-8de7ce/manifest   # or paste the manifest below

Manifest

An Open Capability Manifest: the router reads it to know what this does, what it costs and when to pick it.

rjayapra-ai-agent-engineering-template-prompt-safety-rev-8de7ce.ocm.jsonjson

{
  "ocm": "1",
  "id": "rjayapra-ai-agent-engineering-template-prompt-safety-rev-8de7ce",
  "kind": "agent",
  "name": "prompt-safety-reviewer",
  "description": "Review prompts and grounding logic for safety, instruction hierarchy, and prompt injection resilience.",
  "publisher": "rjayapra",
  "version": "1.0.0",
  "capabilities": {
    "domains": [
      "general"
    ],
    "tags": [
      "agent-md",
      "github-custom-agents"
    ],
    "languages": [
      "en"
    ]
  },
  "quality_prior": 0.6,
  "examples": [
    "Review prompts and grounding logic for safety, instruction hierarchy, and prompt injection resilience."
  ],
  "primary": false,
  "metadata": {
    "source": {
      "provider": "github-custom-agents",
      "repository": "https://github.com/rjayapra/ai-agent-engineering-template",
      "path": ".github/agents/prompt-safety-reviewer.agent.md",
      "ref": "463aedb98aa5057904fda311cc8765b09fb991ed",
      "url": "https://github.com/rjayapra/ai-agent-engineering-template/blob/463aedb98aa5057904fda311cc8765b09fb991ed/.github/agents/prompt-safety-reviewer.agent.md",
      "key": "rjayapra/ai-agent-engineering-template/.github/agents/prompt-safety-reviewer.agent.md"
    }
  },
  "instructions": "# Prompt Safety Reviewer Agent\n\n## Purpose\n\nReview prompts and grounding logic for safety, instruction hierarchy, and prompt injection resilience.\n\n## Use when\n\n- Files under `prompts/` change.\n- Retrieval, grounding, tool use, or system prompt behavior changes.\n\n## Instructions\n\nReview using:\n\n- `prompts/system-prompt.md`\n- `prompts/grounding-instructions.md`\n- `prompts/safety-instructions.md`\n- `tests/prompt-injection-tests.md`\n- `docs/ai-evaluation.md`\n\nCheck for:\n\n- retrieved content overriding system instructions\n- unclear refusal rules\n- missing fallback behavior\n- unsafe tool/action per",
  "cost": {
    "context_tokens": 206
  }
}

Fetch it by URL: GET /api/v1/registry/rjayapra-ai-agent-engineering-template-prompt-safety-rev-8de7ce/manifest?version=1.0.0