What REA “evidence” means

Last updated: · By the RepoRead editorial team · Unofficial guide
Quick answer

In REA, Evidence is the structured record behind a finding: artifact identity, source locations, observations, and limitations / unknowns. Agents and the rea CLI share the same Evidence contracts. Chat text from your model is not Evidence.

REA 6.5.0 (October 14, 2026) documents Evidence heavily in docs/cli.md. RepoRead is an unofficial explainer.

Quick picture

Target (authorized) → REA workflow → Evidence (+ unknowns)
                              ↓
                     inspect-analysis-view / agent tools
                              ↓
                     Summary or paged “view” for humans
  • Evidence: authenticated observations with locations and coverage limits.
  • View: a smaller projection (summary, page, managed collection) over that Evidence.
  • Agent answer: optional narration — verify against Evidence before you trust it.

Facts

ConceptEvidence (structured analysis record)
CLI helpersSave --json, then rea inspect-analysis-view
Import/exportrea evidence-import / rea evidence-export / rea compare (CLI docs)
Example size~334 MB Obsidian Evidence → ~10 KB summary view (upstream)
Verified2026-10-18 (UTC)

Save and project (JavaScript example)

rea analyze-javascript-application /absolute/path/to/app --json > app-evidence.json
jq -c '{source: {kind: "inline", evidence: .}, view: {kind: "summary"}}' app-evidence.json > app-view.json
rea inspect-analysis-view app-view.json

Module page:

{"kind": "page", "collection": "modules", "offset": 0, "limit": 32}

Retained Evidence (MCP / large results)

Managed and other workflows may return a parent_evidence_id. Page collections with a retained-evidence source, for example:

{
  "source": { "kind": "retained-evidence", "evidence_id": "<parent_evidence_id>" },
  "view": { "kind": "managed", "collection": "attributes", "offset": 0, "limit": 16 }
}

Exact collection names depend on the workflow — see the managed-code and MCP contract docs.

Import, export, compare

From the CLI Evidence section:

rea evidence-import /absolute/path/to/evidence/bundle.json
rea evidence-export /absolute/path/to/evidence/bundle.json /absolute/path/to/evidence/canonical.json
rea compare /absolute/path/to/evidence/left.json /absolute/path/to/evidence/right.json

Use these for audit trails and build-to-build diffs — not as a substitute for authorization checks.

How to read Evidence responsibly

DoDon’t
Read unknowns and coverage fieldsTreat a confident chat summary as ground truth
Keep absolute paths and digests with notesPaste proprietary Evidence into public issues
Re-run after rea update when formats changeAssume absence of a field means absence of a feature
Analyze authorized targets onlyUse REA to bypass DRM or licenses

FAQ

Is Evidence the same as the agent’s chat reply?

No. Evidence is the structured analysis record REA returns. The agent’s prose is a separate interpretation — always prefer the Evidence (and its unknowns) when something matters.

Why save Evidence to a file?

Complete application Evidence can be hundreds of megabytes. Upstream’s Obsidian example is ~334 MB raw → ~10 KB summary via inspect-analysis-view.

What is a “view”?

A projection over saved or retained Evidence (summary, module page, managed collections, etc.) so humans and agents can read a slice without reloading everything.

Does exit code 0 mean the analysis is complete and perfect?

Exit 0 means the operation completed; the result may still include partial evidence, warnings or unresolved questions (CLI docs).

Can Evidence prove a binary is safe?

No. Evidence documents what was observed under stated limits. Security conclusions need human review and a proper assessment process.

Sources