HI Fabio,

Similar to JB, this project looks interesting. In a different area we
have Apache Sourcelume working upon Dataset provenance for LLM
training. This project may be of interest to the larger Responsible AI
initiative at Apache (https://rai.apache.org/). I'll share some links
to the community.

--Jamie

On Tue, Nov 10, 2026 at 1:16 PM Fabio Salvadori via general
<[email protected]> wrote:
>
> Hi JB,
> Thanks for taking a look. Yes, we have a few people already contributing. 
> Vedanth has worked on conformance in both the Python and TypeScript repos, 
> Alejandro Florez has contributed conformance cases and structured CLI output, 
> and Joan Abad recently reproduced an issue in the MCP guard and opened a fix. 
> I’m still the sole maintainer. TypeScript conformance PR 
> [https://github.com/pic-standard/pic-standard-ts/pull/21] · MCP guard fix 
> [https://github.com/pic-standard/pic-standard/pull/201]
> There’s been some hands-on work with other projects too.
> AgentRust independently reproduced our MCP guard integration and moved it to 
> their Verified tier. Separately, collaborators working on APS, Probity, and 
> HorizonShield have exercised PIC’s verifier in a shared refund sandbox. Both 
> are integration experiments at this stage. AgentRust verification 
> [https://github.com/agentrust-io/integrations/pull/398] · Refund sandbox 
> discussion [https://github.com/aeoess/agent-governance-vocabulary/issues/193]
>
> Cordum’s maintainers reviewed and merged a PIC verification gate into their 
> packs repo and made follow-up changes there. Arjun Garg and I used PIC’s 
> signed-approval verifier in his public MCP retry experiment to test what 
> happens when a response is lost or a signing key is revoked. HERMETICUM also 
> independently stress-tested an earlier release, and we turned five of their 
> findings into regression tests. Cordum pack 
> [https://github.com/cordum-io/cordum-packs/pull/183] · MCP experiment 
> [https://github.com/arjun2075/mcp-b1-lost-response-repro/pull/2] · 
> Stress-test follow-up 
> [https://github.com/pic-standard/pic-standard/blob/main/docs/independent-reviews/hermeticum-v0.8.3.md]
> My next step is to ask returning contributors if they’d like to help review 
> changes and triage issues through the v1.0 cycle. Conformance across the two 
> implementations is the natural place to start because there’s already someone 
> working in both. I’d also like to use the existing integrations to get 
> feedback from people trying PIC in their own workflows and turn that into 
> issues they can help shape.
> How much shared maintenance would you want to see in practice before we start 
> putting together a proposal?
> Thanks,
> Fabio
> On 11/10/2026 10:35:03 AM, Jean-Baptiste Onofré <[email protected]> wrote:
> Hi Fabio
>
> Thanks for sharing! Even if it's not a domain I'm super familiar with, it 
> looks interesting.
>
> Can you share some details about your plan regarding community growth? Do you 
> already have potential contributors or players interested?
>
> Regards
> JB
>
> On Mon, Nov 9, 2026 at 7:48 PM Fabio Salvadori via general 
> <[email protected] [mailto:[email protected]]> wrote:
>
> Hi everyone,
> I’m Fabio Marcello Salvadori, the creator and maintainer of PIC Standard 
> (Provenance & Intent Contracts). I built PIC to answer a question that comes 
> up once AI agents can act on real systems: when an agent is about to move 
> money, export sensitive data, or delete something, how can the system verify 
> the evidence for that exact action before it runs?
> PIC defines an open contract that carries the proposed tool call together 
> with its intent, provenance, claims, and supporting evidence. A verifier can 
> run locally with keys the operator controls. In our signed-refund example, a 
> merchant approves a €40 refund on a specific payment. The approval is bound 
> to the exact call. Change the payment, amount, or currency and the tool never 
> runs. The example tests 22 cases and checks what actually reaches the tool.
> PIC is Apache-2.0 licensed, with a Python reference verifier, a separate 
> TypeScript implementation, and a public suite of 91 conformance vectors. CI 
> compares Python and TypeScript on the cases both implement. External 
> contributors have added conformance cases and tooling, and findings from an 
> independent protocol stress test have become regression tests.
> I’d like PIC to become a shared contract that agent frameworks, tool 
> gateways, and policy systems can implement under open governance. That’s why 
> I’m exploring Apache as a long-term home. I’d welcome your view on whether 
> PIC fits the Incubator and what you’d want to see in a proposal.
>
> If someone is interested in helping shape that proposal as a Champion, I’d be 
> glad to talk.
> Project: https://github.com/pic-standard/pic-standard 
> [https://github.com/pic-standard/pic-standard] 
> [https://github.com/pic-standard/pic-standard 
> [https://github.com/pic-standard/pic-standard]]
> Signed-refund example: 
> https://github.com/pic-standard/pic-standard/tree/main/examples/refund-approval
>  
> [https://github.com/pic-standard/pic-standard/tree/main/examples/refund-approval]
>  
> [https://github.com/pic-standard/pic-standard/tree/main/examples/refund-approval
>  
> [https://github.com/pic-standard/pic-standard/tree/main/examples/refund-approval]]
> Thanks,
>
> Fabio Marcello Salvadori
> [6f0ab60c-c26c-4943-83fb-7b4bccfe357e]
> [f3cde913-3cc6-432b-b4bc-b9b14e940fa8]

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to