# AI use cases for Agent Safe Pipeline (decionis/agent-safe-pipeline) Repo: [decionis/agent-safe-pipeline](https://github.com/decionis/agent-safe-pipeline) | Use cases: 4 | Page: https://dedupedai.com/use-cases/decionis-agent-safe-pipeline/ AgentSafe is an open-source gateway that sits in front of an AI agent's or API's consequential HTTP requests, checks each one for authorization, and then forwards it once, holds it for a person, or refuses it, leaving chained evidence of each decision. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Test whether an agent can reach a payments API directly: https://dedupedai.com/use-cases/decionis-agent-safe-pipeline/test-agent-boundary-before-install/ - Hold high-value API calls until a person approves them: https://dedupedai.com/use-cases/decionis-agent-safe-pipeline/hold-high-value-api-calls-for-approval/ - Keep a chained record of every agent API action: https://dedupedai.com/use-cases/decionis-agent-safe-pipeline/record-chained-evidence-of-agent-actions/ - Preview what enforcement would change on live agent traffic: https://dedupedai.com/use-cases/decionis-agent-safe-pipeline/preview-enforcement-on-live-agent-traffic/ ## How to test whether an agent can reach a payments API directly with Agent Safe Pipeline Repo: [decionis/agent-safe-pipeline](https://github.com/decionis/agent-safe-pipeline) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Compare direct, shadow and enforced requests before any real traffic goes through a gateway. **Who it is for:** A platform engineer deciding whether an AI agent with a valid API credential should be allowed to call a consequential endpoint. **The problem:** A valid credential does not show that a specific request was authorized, and the target service accepts whatever arrives with a real credential. **The outcome:** A table showing which adversarial requests reached the target directly, in shadow and under enforcement, with a verdict and an exit code. ### Steps 1. Install AgentSafe with Homebrew: `brew tap decionis/agent-safe https://github.com/decionis/agent-safe-pipeline && brew trust decionis/agent-safe && brew install agentsafe` 2. Run the boundary test: `agentsafe test` 3. Read the Exposure line and the Verdict; exit code 0 means the boundary holds and 1 means exposure. 4. Start the gateway in shadow mode against your API, using the next-step line the test prints: `agentsafe proxy --upstream --mode shadow` (inferred, check the README) 5. Compare what shadow mode would have held or refused with what the direct column reached. ### Prompt to paste into your agent ```text Using the decionis/agent-safe-pipeline repository, install AgentSafe and run `agentsafe test`. Report the Exposure line and the Verdict for the boundary test against if you have one, otherwise against the built-in synthetic target. Confirm the exit code is 0 before you say the boundary holds. ``` ### Caveat Without a Decionis key the test uses a local demo authority with a synthetic policy, so the result shows how the boundary behaves, not how your real policy would decide. 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/decionis-agent-safe-pipeline/test-agent-boundary-before-install/ ## How to hold high-value API calls until a person approves them with Agent Safe Pipeline Repo: [decionis/agent-safe-pipeline](https://github.com/decionis/agent-safe-pipeline) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Route consequential HTTP requests through a gateway that holds the risky ones for approval. **Who it is for:** An operations lead running an AI agent that calls a payments or customer-records API, who wants a person to approve actions above a certain level. **The problem:** Consequential requests from an agent should not reach the target without review, but small routine requests should not wait on a person. **The outcome:** A large payment returns 202 and is held without reaching the upstream, while a small payment is forwarded once with its dossier id. ### Steps 1. Install AgentSafe with Homebrew: `brew tap decionis/agent-safe https://github.com/decionis/agent-safe-pipeline && brew trust decionis/agent-safe && brew install agentsafe` 2. Start the gateway in front of your API: `agentsafe proxy --upstream http://localhost:3000 --port 8080` 3. Send a large payment through the gateway: `curl -i -X POST http://127.0.0.1:8080/payments -H 'content-type: application/json' -d '{"amount": 500}'` 4. Check that the response shows Execution HELD and that the upstream received nothing. 5. Point the agent's HTTP base URL at http://127.0.0.1:8080 instead of the target service. (inferred, check the README) ### Prompt to paste into your agent ```text Using the decionis/agent-safe-pipeline repository, install AgentSafe and start `agentsafe proxy --upstream --port 8080`. Send one POST request to /payments through the gateway with an amount above the approval threshold, and report the decision and the HTTP status. Verify that the upstream received no request for that call. ``` ### Caveat The approval step uses a loopback double that simulates the Presence ceremony for examples and proves nothing about a real human approval. 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/decionis-agent-safe-pipeline/hold-high-value-api-calls-for-approval/ ## How to keep a chained record of every agent API action with Agent Safe Pipeline Repo: [decionis/agent-safe-pipeline](https://github.com/decionis/agent-safe-pipeline) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Print and store the evidence lines for each intent, decision and forward the gateway handles. **Who it is for:** A security or compliance reviewer who needs to show what an AI agent proposed, what was decided and what was forwarded. **The problem:** Without a record tied to each decision, it is hard to reconstruct what an agent tried to do and why a request was allowed or refused. **The outcome:** Verbose output with chained evidence lines for each request, plus a Decision Dossier id beside each forwarded or refused action. ### Steps 1. Install AgentSafe with Homebrew: `brew tap decionis/agent-safe https://github.com/decionis/agent-safe-pipeline && brew trust decionis/agent-safe && brew install agentsafe` 2. Check what would stop the gateway from governing: `agentsafe doctor` 3. Start the gateway with verbose evidence: `agentsafe proxy --upstream http://localhost:3000 --port 8080 --verbose` (inferred, check the README) 4. Send a small payment that should be allowed: `curl -i -X POST http://127.0.0.1:8080/payments -H 'content-type: application/json' -d '{"amount": 50}'` (inferred, check the README) 5. Read the chained evidence lines and note the dossier id returned with the decision. ### Prompt to paste into your agent ```text Using the decionis/agent-safe-pipeline repository, install AgentSafe, run `agentsafe doctor`, and start `agentsafe proxy --upstream --port 8080 --verbose`. Send one allowed request and one blocked request through the gateway, then list the evidence lines and dossier ids for both. Verify that each request has its own evidence line before you report back. ``` ### Caveat By default the quickstart does not write evidence to disk, so use --verbose or set evidence.journalDir to keep a record, and the dossiers are synthetic unless a Decionis key is connected. 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/decionis-agent-safe-pipeline/record-chained-evidence-of-agent-actions/ ## How to preview what enforcement would change on live agent traffic with Agent Safe Pipeline Repo: [decionis/agent-safe-pipeline](https://github.com/decionis/agent-safe-pipeline) | Difficulty: Intermediate | Time: 1 hour | Basis: From the README Run the gateway in shadow mode and read the report of what enforcement would have held or refused. **Who it is for:** A team lead who wants evidence from real agent traffic before switching a gateway from shadow to enforcement. **The problem:** Turning on enforcement without data risks blocking legitimate work, and leaving it off leaves the target exposed. **The outcome:** A shadow report by action showing what enforcement would have held or refused, and how many of those the upstream accepted as sent. ### Steps 1. Install AgentSafe with Homebrew: `brew tap decionis/agent-safe https://github.com/decionis/agent-safe-pipeline && brew trust decionis/agent-safe && brew install agentsafe` 2. Start the gateway in shadow mode in front of your agent's API: `agentsafe proxy --upstream --mode shadow` (inferred, check the README) 3. Let the agent run its normal workload through the gateway for a representative period. 4. Read the shadow report: `agentsafe status` 5. Review the held and refused actions by type, then decide whether to switch to enforcement mode. ### Prompt to paste into your agent ```text Using the decionis/agent-safe-pipeline repository, install AgentSafe and run it in shadow mode in front of . After the agent has run its normal tasks, run `agentsafe status` and summarize what enforcement would have held or refused, by action. Verify that the shadow report reflects requests the agent actually sent before recommending enforcement. ``` ### Caveat Shadow results depend on the authority in use; with no Decionis key the report comes from a local demo policy, not your organization's real policy. 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/decionis-agent-safe-pipeline/preview-enforcement-on-live-agent-traffic/