Governed RMF automation

awaiting approvaltieredInteractive simulation

Preview of the production 11-step RMF DAG with cryptographic enforcement receipts. Showing 5 of 11 lifecycle steps — one parked awaiting human approval. This is the governed multi-agent orchestrator story, not a chatbot demo.

Information system

DEMO-IS-ALPHA (Synthetic Boundary)

Run cost

$0.51

Governance

Awaiting human approval

cATO target

yes

Budget

$0.51 / $5.00

RMF lifecycle DAG

5 of 11 steps visible · all governed

Full production DAG continues through Generate POA&Ms → Collect Evidence → Monitor → Remediate → Authorize (AO-only) → Keep cATO Green.

Governed step log

Per-step status, stakes, and approval

01Registergoverned
completed
Information system registered under boundary DEMO-IS-ALPHA. Ownership + mission mapped.
Receipt
linked
Cost
$0.04
Provenance
1 stamp
02Categorizegoverned
completed
FIPS 199 categorization proposed: Confidentiality Moderate, Integrity Moderate, Availability Low.
Receipt
linked
Cost
$0.11
Provenance
2 stamps
03Baselinegoverned
completed
NIST 800-53 Rev 5 Moderate baseline selected; overlay set: CNSSI 1253 + org tailoring draft.
Receipt
linked
Cost
$0.09
Provenance
1 stamp
04Select Controlsgovernedhigh-stakes
awaiting approval
Agent proposes 287 controls (baseline 261 + 26 tailored). 4 high-stakes controls flagged for AO review: AC-6, AU-6, CM-6, SI-4.
Receipt
linked
Cost
$0.27
Provenance
3 stamps
This step is parked and requires human approval before the orchestrator continues. High-stakes control selection is never auto-committed.
Agent proposal ready for review
05Assessgovernedhigh-stakes
pending
Queued — waits for Select Controls approval before assessment planning starts.
Receipt
—
Cost
$0.00
Provenance
—

Enforcement receipt chain

Signed, append-only policy decisions

Receipts

Allow

Deny

Every model invocation is policy-checked. Allow and deny decisions append to a signed, hash-chained receipt ledger — independently verifiable offline.

Head: d4f6b20c5e…d3c6b8

#4deny
d4f6b20c5e1a37…f1d3c6b8
prev → c3e5a19b…b5a7
High-stakes control selection requires human approval — step parked
rmf-executor·select-controls·enforcement-ed25519-v3
#3allow
c3e5a19b4d0f26…e0c2b5a7
prev → b2d4f08a…a4f6
Baseline selection matches org catalog allow-list
rmf-analyst·baseline·enforcement-ed25519-v3
#2allow
b2d4f08a3c9e15…d9b1a4f6
prev → a7c3e91f…f3e5
Categorization proposal within policy; provenance attached
rmf-analyst·categorize·enforcement-ed25519-v3
#1allow
a7c3e91f2b8d04…c8a0f3e5
Within per-agent budget; low-stakes step; kill-switch clear
rmf-planner·register·enforcement-ed25519-v3

Run governance

Autonomy modetiered
AO decisionhuman-only
Kill-switcharmed
Run idrmf-run-demo-7c4e2a91