Agent EvidencePortable trust records for agent work.

Make answers, artifacts, tool decisions, reviews, redactions, and replays auditable without becoming the runtime, UI, trace backend, or storage system.

Claim-level grounding

Connect every important assertion to supporting, contradicting, qualifying, or background evidence.

Portable provenance

Preserve source, tool, model, artifact, human, peer-agent, runtime, trace, and span refs as a production graph.

Verification is not review

Keep automated checks, human verdicts, waivers, rubrics, sign-off facts, and open issues separate but linked.

Replay honesty

Describe what can be reconstructed, what is approximate, and which facts are missing, expired, redacted, or unavailable.

Redaction-aware export

Share evidence packs with stable ids, schemas, hashes, access metadata, and explicit redaction records.

AI-friendly docs

Publish llms.txt, llms-full.txt, JSON Schemas, examples, and versioned snapshots for agent consumption.

What Agent Evidence Defines

Contract	What it answers
Evidence pack	What scoped outcome is being audited, and how complete is the evidence?
Claim map	Which claims were made, and how is each claim supported, contradicted, qualified, or left unsupported?
Source map	Which sources were selected, omitted, stale, contradicted, redacted, or unavailable?
Provenance chain	Which tools, models, artifacts, humans, peers, runtime events, traces, and spans produced the outcome?
Verification and review	Which checks ran, and who or what issued a verdict?
Replay and export	What can be reconstructed, and what is safe to share?

Start Here

For AI Clients

llms.txt: concise navigation index.
llms-full.txt: current English core documentation in one file.
llm.txt and llm-full.txt: compatibility aliases.

Agent Standards Ecosystem

Evidence owns trust, review, replay, redaction, and export records. Runtime owns execution, UI owns projection, Knowledge owns source-grounded context, and Artifact systems own durable deliverables.