Public-sector auditability profile for verifiable AI agents
Maintainers usually reply within 1 day
@fernandacamiladecarli is already working on this.
Since Sep 7, 2026.
Assessment
This issue has not been assessed yet.
Description
Spec section affected
§3 Trust Record fields; §4 Verification / conformance guidance; potentially an informative annex for sector-specific profiles.
Problem
The current TRACE specification appears to focus on general evidence records for verifiable AI agent activity, but it does not yet define a public-sector auditability profile for agents used in procurement, public oversight, or administrative decision-support.
For public institutions, auditability requires more than a record that an agent acted. Auditors and oversight bodies may need evidence of:
agent identity;
delegated authority;
authorized tools;
data access;
human approval or override;
policy exceptions;
incident or escalation triggers;
retention requirements for later review.
Without a public-sector profile, different agencies or vendors may implement evidence records inconsistently, making later audit, accountability, and oversight harder.
Proposed change
Add an optional, non-breaking public-sector auditability profile to TRACE, either as an informative annex or sector-specific guidance.
Proposed normative/informative text:
Implementations MAY support a public-sector auditability profile for AI agents used in procurement, oversight, regulatory, or administrative decision-support contexts.
When this profile is used, Trust Records SHOULD include sufficient evidence to support later institutional review of:
the identity and version of the agent;
the public authority, delegation, or approval under which the agent acted;
the tools, systems, or external services accessed by the agent;
the categories of data accessed or processed;
any human approval, override, denial, or escalation;
any policy exception or rule violation;
incident, near-incident, or escalation triggers;
retention metadata necessary for audit or oversight review.
This profile SHOULD NOT require disclosure of protected, confidential, classified, or legally restricted information in the Trust Record itself. Where appropriate, the Trust Record MAY reference protected evidence through controlled-access identifiers, hashes, or other mechanisms that preserve verifiability without exposing sensitive content.
This profile is intended to support public accountability, auditability, and institutional oversight without changing the core TRACE wire format.
Backward compatibility
Non-breaking (new optional field, informative addition)
Breaking (removes or changes required field, changes wire format, changes algorithm set)
Motivation
Public institutions are increasingly exploring AI agents for procurement, auditing, regulatory, and administrative workflows. In these settings, accountability depends not only on technical performance, but also on whether later reviewers can verify who authorized the agent, what it did, what data and tools it used, whether a human approved or overrode the action, and whether exceptions or incidents were escalated.
A public-sector auditability profile would help agencies, vendors, auditors, and oversight bodies use TRACE in a more consistent way. It could also connect TRACE with adjacent AgenTrust work, especially Agent Manifest fields related to identity, delegated authority, tools, policies, provenance, and human approvals.
This proposal could connect to Agent Manifest fields, but I am opening it here because the main contribution would be a TRACE profile for public-sector auditability and evidence events.
My background is in public oversight and procurement accountability, not core infrastructure engineering. I am interested in whether this kind of governance-focused profile would be useful for the AgenTrust ecosystem.
Related issues or PRs
None yet.
- Dominant language
- Python
- Stars
- 37
- Forks
- 33
- Avg merge
- 1d 4h
- Merged PRs (30d)
- 122
Getting set up
- No Dockerfile or Docker Compose file
- Has a pull request template
- Read the contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from agentrust-io/trace-spec
-
Difficulty 5/5 Over a week Newbie friendliness 25/100
agentrust-io/trace-spec#593 · 1 comment ·
Maintainers usually reply within 1 day
-
bug needs-review:HIGH
Difficulty 5/5 Over a week Newbie friendliness 20/100
agentrust-io/trace-spec#591 · 4 comments ·
Maintainers usually reply within 1 day
-
prototype(manifest_requirements): resolve accepted_methods without duplicating bindings and reject unconfigured binding endpoints with ProfileErrorPossibly taken @drk-ram claimed this 1 day ago. Open
Difficulty 3/5 1-2 days Newbie friendliness 78/100
agentrust-io/trace-spec#580 ·
Maintainers usually reply within 1 day
-
feat(delegation): add parent_record_hash, verify_delegation_link, and verify_delegation_chain to agentrust_tracePossibly taken @drk-ram claimed this 1 day ago. Open
Difficulty 4/5 3-5 days Newbie friendliness 58/100
agentrust-io/trace-spec#581 ·
Maintainers usually reply within 1 day
-
revocation,trust-record-verify: implement es256 and es384 bundle verification and add EC revocation vectors (#448 items 1, 10)Possibly taken @drk-ram claimed this 1 day ago. Open
Difficulty 4/5 3-5 days Newbie friendliness 52/100
agentrust-io/trace-spec#584 ·
Maintainers usually reply within 1 day
All issues in agentrust-io/trace-spec
Similar issues
-
Difficulty 1/5 Under an hour Newbie friendliness 80/100
521xueweihan/HelloGitHub#4063 ·
-
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
-
Difficulty 2/5 1-3 hours Newbie friendliness 62/100
Vanilla-OS/apx-gui#112 ·
-
Devstral and Voxtral: section 5 fails in us-east-1 with 400 "Model does not support audio modality"Open
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
aws-samples/sample-per-model-bedrock#409 ·
Maintainers usually reply within 1 day
-
DuckDB 2.0 breaks to_pyarrow_batchesPossibly taken @DG47 claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
ibis-project/ibis#12196 ·
Maintainers usually reply within 1 day