Sample proof run
AI Agent Action Proof Run
This sample shows how another party checks a proof bundle after an AI agent acts. Inspect the action boundary, evidence manifest, receipt, verifier result, challenge path, and digest manifest before requesting a proof run for your own workflow.
Three-minute buyer walkthrough
Can another party inspect who approved the action, what ran, what evidence survived, and what the verifier checked?
A CISO, AI platform owner, GRC lead, or auditor needs to read the bundle without guessing what each file means.
The sample demonstrates receipt shape and verifier path. It is not a production deployment, legal compliance claim, or whole-program assurance claim.
Sample outcome
What this sample shows before you open the artifacts
A sample AI agent performed one bounded code or configuration workflow after human approval.
VERIFY_RESULT.json reports pass_with_sample_limitations for the sample receipt shape, manifest checks, and declared evidence boundaries.
This sample does not prove production deployment, legal compliance, source-system truth, production signing-key custody, or complete AI governance.
You can inspect how authority, action scope, evidence, receipt, verifier result, challenge path, and manifest fit together before requesting a proof run for your own workflow.
Read the buyer walkthrough first if you want the plain-English path. Open the artifacts when you want to inspect the bundle directly.
How to inspect this sample
Confirm the sample is scoped to one workflow, one agent/tool path, and one touched system.
ACTION_BOUNDARY.json
Inspect who may approve, run, review, and challenge the action.
AUTHORITY_MAP.json
Review captured artifacts, hashes, sources, and declared evidence gaps.
EVIDENCE_MANIFEST.json
See how approval, action, evidence, result, and limits are bound into one record.
RECEIPT.json
See what passed, what failed, and what remained outside the verifier boundary.
VERIFY_RESULT.json
Understand how another party can inspect, dispute, or ask for stronger evidence.
CHALLENGE_PATH.md
Check the published file digests so artifact drift is visible.
MANIFEST.sha256
The manifest is the sample drift check, not a production custody claim.
AI agent proposes and performs a bounded code or configuration change after human approval.
Can another party inspect who approved the action, what ran, what system it touched, what evidence survived, and what could not be verified?
Use the sample to inspect the proof shape, then submit one bounded agent-assisted workflow for your own proof run.
Artifact set
The one workflow, one action path, and one system boundary under review.
Open artifact
Who can approve, run, review, and challenge the agent-assisted action.
Open artifact
The captured artifacts, hashes, sources, and known evidence gaps.
Open artifact
The signed or simulated receipt binding approval, action, evidence, result, and limits.
Open artifact
The verifier result showing pass, fail, or limits for the sample bundle.
Open artifact
How a third party can inspect, dispute, or request stronger evidence.
Open artifact
Digest list for checking that the sample artifacts did not silently drift.
Open artifact
Proof-run shape
- one workflow
- one action path
- one receipt
- one verifier result
- one challenge path
Boundary
This sample proves the receipt shape and verifier path only. It does not claim production deployment, legal compliance, or complete AI governance coverage.
Ready to test your own workflow?
Use the sample to inspect the proof shape before submitting your own workflow. The request page asks for the workflow name, agent/tool involved, system touched, approval boundary, evidence available, buyer email, and urgency.