# AI use cases for Dodoguard (dodoguardai/dodoguard) Repo: [dodoguardai/dodoguard](https://github.com/dodoguardai/dodoguard) | Use cases: 4 | Page: https://dedupedai.com/use-cases/dodoguardai-dodoguard/ DodoGuard is a self-hosted tool that evaluates, red-teams, and audits AI agents and agent apps before and after deployment, with reports you keep on your own infrastructure. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Run a factuality eval on a prompt before release: https://dedupedai.com/use-cases/dodoguardai-dodoguard/eval-llm-prompt-before-release/ - Red-team a local model for prompt injection: https://dedupedai.com/use-cases/dodoguardai-dodoguard/red-team-local-model-with-ollama/ - Audit a Dify or Coze agent for misconfig and attack surface: https://dedupedai.com/use-cases/dodoguardai-dodoguard/check-dify-coze-fastgpt-agents-for-misconfig/ - Export release evidence to support a go-live decision: https://dedupedai.com/use-cases/dodoguardai-dodoguard/export-release-evidence-for-go-live-decision/ ## How to run a factuality eval on a prompt before release with Dodoguard Repo: [dodoguardai/dodoguard](https://github.com/dodoguardai/dodoguard) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Score an LLM prompt against test cases from the command line in about one minute of run time. **Who it is for:** A developer who changed a prompt or switched models and needs to confirm answer quality did not drop. **The problem:** Changing a prompt or model can quietly lower answer quality, and spot-checking chat does not catch it. **The outcome:** A pass or fail result for each test case, so a prompt or model change can be checked the same way every time. ### Steps 1. Run `git clone https://github.com/dodoguardai/dodoguard.git` 2. Run `cd dodoguard && npm ci && npm run build` 3. Run `cp -R examples/openai-eval-factuality ./my-eval && cd my-eval` 4. Set your key with `export OPENAI_API_KEY=" "`, replacing the space with your real key 5. Run `../node_modules/.bin/dodoguard eval` ### Prompt to paste into your agent ```text In the dodoguard repo, copy examples/openai-eval-factuality to a new folder, set OPENAI_API_KEY, and run ../node_modules/.bin/dodoguard eval from that folder. Report the pass and fail count for each test case. Verify that the run finished and wrote a result for every case before you report success. ``` ### Caveat The quick start uses OpenAI, and the README lists other providers, so check the example config before you switch providers. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/dodoguardai-dodoguard/eval-llm-prompt-before-release/ ## How to red-team a local model for prompt injection with Dodoguard Repo: [dodoguardai/dodoguard](https://github.com/dodoguardai/dodoguard) | Difficulty: Advanced | Time: 45 minutes | Basis: From the README Run adversarial probes against a model hosted with Ollama to find jailbreaks and guardrail bypasses. **Who it is for:** A security-minded engineer who wants to test a locally hosted model without sending prompts to a cloud provider. **The problem:** System prompts and guardrails can be bypassed, and teams need to find that before users do. **The outcome:** A list of red-team findings for the local model, with evidence you can keep private. ### Steps 1. Install and start Ollama on your machine and pull the model you want to test 2. Clone the repo and run `npm ci && npm run build` from the Dodoguard folder as in the quick start 3. Open the examples/redteam-ollama folder in the repo and read its config 4. Point the example config at your local Ollama model and the system prompt you want to test (inferred, check the README) 5. Run the red-team example with the Dodoguard CLI from the examples/redteam-ollama folder and review the findings ### Prompt to paste into your agent ```text In the dodoguard repo, use the examples/redteam-ollama config to red-team a local Ollama model named . Return the list of findings with severity and evidence. Verify that each finding links to the prompt that caused it. ``` ### Caveat The README only links to the Ollama example without listing its exact commands, so confirm the run steps in that folder before you rely on them. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/dodoguardai-dodoguard/red-team-local-model-with-ollama/ ## How to audit a Dify or Coze agent for misconfig and attack surface with Dodoguard Repo: [dodoguardai/dodoguard](https://github.com/dodoguardai/dodoguard) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Scan an agent app built on a hosted platform and get findings with severity and a remediation path. **Who it is for:** A platform owner who builds agents in Dify, Coze, FastGPT, or n8n and must confirm the app is configured safely before go-live. **The problem:** Agent apps can ship with misconfigurations and exposed attack surface that a demo never shows. **The outcome:** A versioned detection task with findings, severity, and evidence for the agent app, plus a list of fixes to retest. ### Steps 1. Start the web console with `docker compose up -d --build` 2. Open http://localhost:3000 in your browser 3. Register the agent app as an asset in the asset ledger 4. Create a detection task for platform checks against that asset 5. Review the findings, fix them, then rerun the task before you decide on go-live ### Prompt to paste into your agent ```text Using the dodoguard web console at http://localhost:3000, register the agent app as an asset and create a platform check task for it. Return the findings with severity and evidence. Verify that the task was rerun after fixes and that no open high-severity finding remains before you recommend release. ``` ### Caveat The README names Dify, Coze, FastGPT, and n8n with Dify-class scans, so check that your platform is covered before you rely on the results. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/dodoguardai-dodoguard/check-dify-coze-fastgpt-agents-for-misconfig/ ## How to export release evidence to support a go-live decision with Dodoguard Repo: [dodoguardai/dodoguard](https://github.com/dodoguardai/dodoguard) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Track an agent through intake, detection, and observation, then export reports you keep private. **Who it is for:** A release owner who must show evidence of testing before an AI agent goes live. **The problem:** Teams often ship agents with no durable findings or reports to show what was tested. **The outcome:** A set of exported reports for one agent that document the findings, the fixes, and the retest before go-live. ### Steps 1. Start the web console with `docker compose up -d --build` 2. Open http://localhost:3000 and register the agent in the asset ledger 3. Run evaluation, red-team, and platform detection tasks against the agent 4. Work through findings in the remediation loop and rerun the tasks 5. Export the reports and keep them in your own infrastructure ### Prompt to paste into your agent ```text Using the dodoguard repo, register the agent in the web console, run its detection tasks, and export the reports. Return the exported report list and the open findings count. Verify that the acknowledged state did not move the agent to released before the remediation loop passed. ``` ### Caveat The README says acknowledge is not release, so confirm the agent is marked released only after your own sign-off in the console. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/dodoguardai-dodoguard/export-release-evidence-for-go-live-decision/