# AI use cases for Open Work Proof (dengyier/OpenWorkProof) Repo: [dengyier/OpenWorkProof](https://github.com/dengyier/OpenWorkProof) | Use cases: 5 | Page: https://dedupedai.com/use-cases/dengyier-openworkproof/ OpenWorkProof is an open protocol that records who authorized an AI agent's work, what the agent did, how it was verified, and who accepted or rejected it, packaged as signed evidence that third parties can check offline. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Verify an agent work package offline without the vendor: https://dedupedai.com/use-cases/dengyier-openworkproof/verify-agent-work-offline/ - Run the Human Agency example to see reserved actions: https://dedupedai.com/use-cases/dengyier-openworkproof/see-human-decision-gates-in-action/ - Attach a four-question evidence report to pull requests: https://dedupedai.com/use-cases/dengyier-openworkproof/add-evidence-report-to-pull-requests/ - Package an agent delivery case for payer review: https://dedupedai.com/use-cases/dengyier-openworkproof/package-delivery-case-for-settlement-review/ - Build an acceptance bundle for an outside human signer: https://dedupedai.com/use-cases/dengyier-openworkproof/prepare-external-human-acceptance-bundle/ ## How to verify an agent work package offline without the vendor with Open Work Proof Repo: [dengyier/OpenWorkProof](https://github.com/dengyier/OpenWorkProof) | Difficulty: Intermediate | Time: 15 minutes | Basis: From the README Check a signed Surface or Acceptance bundle with the owp CLI, no access to the original platform. **Who it is for:** A client, auditor, or reviewer who received a bundle from an AI service provider and needs to confirm the work chain without access to the provider's systems. **The problem:** A log file or a status message from the agent platform cannot show independently whether the work was authorized, followed its scope, and was accepted. **The outcome:** A clear pass or fail result from the offline verifier for the Surface or Acceptance bundle, based only on the bundle contents. ### Steps 1. Install the package with `python -m pip install openworkproof`. 2. Confirm the CLI is available with `owp --help`. 3. For a Surface Bundle, run `owp surface-verify PATH` with PATH set to the bundle folder. 4. For an Acceptance Bundle, run `owp acceptance-bundle-verify DIRECTORY` with DIRECTORY set to the bundle folder. 5. Record the exit code and report it with the bundle identifier. ### Prompt to paste into your agent ```text Using the OpenWorkProof repo at https://github.com/dengyier/OpenWorkProof, install the openworkproof package with pip and run `owp acceptance-bundle-verify ` on the bundle at . Report the result and the exit code, where ACCEPTED is 0 and REJECTED is 2. Verify that the command ran without reading any private key from the bundle folder. ``` ### Caveat A VERIFIED or ACCEPTED result proves only that the evidence supports that state; the README says it does not prove the business result is correct or that the client has paid. 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/dengyier-openworkproof/verify-agent-work-offline/ ## How to run the Human Agency example to see reserved actions with Open Work Proof Repo: [dengyier/OpenWorkProof](https://github.com/dengyier/OpenWorkProof) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Run the minimal Human Agency Profile script to see which agent actions are allowed and which need a human. **Who it is for:** A developer or technical lead who wants to understand how a person keeps decision rights over an AI agent before adding the protocol to a workflow. **The problem:** It is unclear how a human's chosen limits combine with what the work order and capability grant allow, and what happens when an action is reserved for a person. **The outcome:** Terminal output showing one delegated action as allowed and one reserved action blocked with AGENCY_HUMAN_DECISION_REQUIRED before execution. ### Steps 1. Clone the repo with `git clone https://github.com/dengyier/OpenWorkProof.git`. 2. Enter the folder with `cd OpenWorkProof`. 3. Create a virtual environment with `python -m venv .venv`. 4. Activate it with `source .venv/bin/activate`. 5. Install in editable mode with `python -m pip install -e .`. 6. Run the example with `python examples/human_agency_profile_v01.py`. ### Prompt to paste into your agent ```text Clone https://github.com/dengyier/OpenWorkProof, create a virtual environment, install it in editable mode, and run `python examples/human_agency_profile_v01.py`. Report whether the output shows 'profile verified : True' and that owp.apply_patch returns AGENCY_HUMAN_DECISION_REQUIRED. Verify that the script writes no application files or ledgers, as the README states. ``` ### Caveat The example is a minimal demonstration of the authorization boundary, not a compliance certification, a performance monitor, or a transfer of legal liability. 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/dengyier-openworkproof/see-human-decision-gates-in-action/ ## How to attach a four-question evidence report to pull requests with Open Work Proof Repo: [dengyier/OpenWorkProof](https://github.com/dengyier/OpenWorkProof) | Difficulty: Advanced | Time: 1 hour | Basis: From the README Use the GitHub Action to report what was claimed, what evidence exists, what limits applied, and the conclusion. **Who it is for:** A team that already runs a pull request delivery process and receives code changes from AI agents, and wants a standard verification report on each change. **The problem:** A pull request that says the agent finished and tests passed does not show the authorization, the evidence chain, or whether anyone has accepted the change. **The outcome:** A report on each pull request with the claim checked, the evidence seen, the constraints on execution, and a VERIFIED, REFUTED, or UNKNOWN conclusion. ### Steps 1. Open the integration file at integrations/github/action.yml in the Open Work Proof repo. 2. Add the action to the pull request workflow of the target repository, following the file's inputs. 3. Open a test pull request and read the four-question report it produces. 4. Confirm that the report states whether the change only reached READY_FOR_ACCEPTANCE and does not claim the client accepted or paid. ### Prompt to paste into your agent ```text Review the GitHub Action definition at integrations/github/action.yml in https://github.com/dengyier/OpenWorkProof and add it to the pull request workflow of . Open a test pull request and report the conclusion in the four-question report. Verify that the report does not describe the change as accepted or paid. ``` ### Caveat The README does not list the action's inputs or workflow settings, so the exact configuration must be read from integrations/github/action.yml. 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/dengyier-openworkproof/add-evidence-report-to-pull-requests/ ## How to package an agent delivery case for payer review with Open Work Proof Repo: [dengyier/OpenWorkProof](https://github.com/dengyier/OpenWorkProof) | Difficulty: Advanced | Time: 1 hour | Basis: From the README Initialize, inspect, verify, and export a Verified Agent Delivery case as a third-party review package. **Who it is for:** A service provider that delivers agent work to a client and needs to hand the payer a checkable record before any payment decision. **The problem:** The payer needs to review the delivery facts without relying on the provider's own database or dashboard. **The outcome:** An exported review package with a deterministic digest and integrity manifest, plus a status derived from the real evidence. ### Steps 1. Create the case folder with `owp delivery-case init CASE_DIR`. 2. Re-derive the status from the evidence with `owp delivery-case inspect CASE_DIR`. 3. Check the case with `owp delivery-case verify CASE_DIR`. 4. Export the third-party review package with `owp delivery-case export CASE_DIR --output-directory OUTPUT_DIR`. 5. Send OUTPUT_DIR to the payer, stating that READY_FOR_SETTLEMENT_REVIEW means the evidence is ready for review, not that payment has happened. ### Prompt to paste into your agent ```text Using https://github.com/dengyier/OpenWorkProof, run `owp delivery-case init `, then `owp delivery-case inspect ` and `owp delivery-case verify `, and export with `owp delivery-case export --output-directory `. Report the derived status. Verify that the status comes from the Surface, Acceptance, and Settlement evidence and not from a pre-written conclusion file. ``` ### Caveat The README says READY_FOR_SETTLEMENT_REVIEW and BOUND do not mean payment, client adoption, or legal recognition, and those facts remain not evidenced without outside proof. 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/dengyier-openworkproof/package-delivery-case-for-settlement-review/ ## How to build an acceptance bundle for an outside human signer with Open Work Proof Repo: [dengyier/OpenWorkProof](https://github.com/dengyier/OpenWorkProof) | Difficulty: Advanced | Time: 1 hour | Basis: From the README Generate an Acceptance Bundle from a ledger so an external Acceptor can sign without a private key entering the tool. **Who it is for:** A delivery lead who needs a client's authorized person to accept or reject agent work, with the acceptance kept separate from verification. **The problem:** Verification passing does not mean the client accepted the work, and the acceptor's private key must not be placed in the tooling used to generate evidence. **The outcome:** An Acceptance Bundle directory built from the ledger and surface evidence, ready for an external acceptor to sign and for a third party to verify. ### Steps 1. Build the bundle with `owp acceptance-bundle-build LEDGER SURFACE --evidence-root PATH --output DIRECTORY`, replacing the placeholders with your ledger, surface, evidence path, and output folder. (inferred, check the README) 2. Hand the draft to the external Acceptor, who signs it outside the tool. 3. Commit the signed object through the append-only transaction the README describes as prepare, sign, commit. 4. Verify the result with `owp acceptance-bundle-verify DIRECTORY`. 5. Record ACCEPTED as exit code 0 or REJECTED as exit code 2, and treat REJECTED as a valid final state. ### Prompt to paste into your agent ```text Using https://github.com/dengyier/OpenWorkProof, run `owp acceptance-bundle-build --evidence-root --output ` and then `owp acceptance-bundle-verify `. Report the exit code and whether it is 0, 2, or 4. Verify that no private key or acceptance receipt was generated by the tool itself. ``` ### Caveat The README says VERIFIED is not the same as ACCEPTED, and ACCEPTED is not the same as paid, settled, legal audit, or adoption. Risk, credentials: The Acceptor's private key must stay with the external signer and must never be placed in the tool, the exporter, or any shared folder. 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/dengyier-openworkproof/prepare-external-human-acceptance-bundle/