All gates CLEAR Provider: OpenRouter · free tier
Spend $1.24 / $5.00 3 approvals CE

Goal inbox

Durable · resumes after restart
GOALSTATUSSTAGEAGENTSBUDGETNEXT ACTION
Crown weekly ops briefWaiting approvalDecisionScout → Analyst → Maker → Auditor4/6 callsApprove decision packet
Vendor quote comparison — linensRunningEvidenceScout, Index1/6 callsNone — auto
DJ SAVAG3 October booking sheetRunningDraftMaker3/6 callsNone — auto
WGU capstone outline reviewBlockedEvidenceScout0/6 callsAdd missing source file
Client Room rollout checklistCompleteReceiptMaker, Auditor6/6 callsExport audit packet

Goals persist locally · every state change writes an event · restart-safe resume

Workroom

Crown weekly ops brief · Stage 5 of 6 — Decision
DECISION PACKETRecommend staffing plan for week of Sep 14 Confidence 78%
Weekend demand requires +2 servers Sat and +1 Sun vs. current roster. [1][3]
Linen vendor B is 12% cheaper but has a 5-day lead time — schedule risk for Sat events. [2] · conflicting
Overtime cost stays under the $650 weekly cap if Sat crew caps at 9. [4]
Evidence used4 sources · 2 verified hashes
1 conflicting pair flagged
RisksVendor B lead time
Sat weather (outdoor)
AlternativeKeep vendor A; +$78 cost, zero schedule risk
Auditor noteIndependent pass · no objections
1 caveat on [2]
Autonomy ceiling: AUT-2 · execution requires owner approval ✓ Approved — receipt logged, work dispatched

PIPELINE

1 GoalDONE
2 EvidenceDONE
3 PlanDONE
4 DraftDONE
5 ReviewYOU
6 Action + receiptLOCKED

EVIDENCE COVERAGE

Supported claims4
Weak / conflicting1
Unsupported0
Stale sources0

Agents & fleet

12 seats · 7 active · routes are per-agent
Sc
Scout
Autonomy AUT-1 · tools: read evidence
Route: free-fast · fallback local
Eval: PASS 9/10 · drift 0.02
An
Analyst
Autonomy AUT-1 · tools: read, structure
Route: mid-tier · fallback free
Eval: PASS 8/10
Mk
Maker
Autonomy AUT-2 · tools: draft, calc
Route: strongest model
Eval: PASS 9/10
Au
Auditor
Autonomy AUT-1 · fresh context · decorrelated route
Route: different provider than Maker
Eval: 1 caveat on [2]
Sw
Switchboard
Autonomy AUT-0 · scope keeper · 0 model calls
Deterministic service
Last: kept run in scope
Or
Oracle (computer)
Autonomy AUT-3* · requires owner dispatch
Shared computer queue · free-only
Hold: owner hold active

Passports bind role, tools, route, budget and data boundary · "connected" only after live probe

Receipts

Append-only · exportable audit trail
TIMEACTOREVENTDETAILHASH
20:41:07Makerdraft.checkpointStaffing plan v3 · 2,140 out toka91f…
20:38:52Auditorreview.complete1 caveat · verdict: PASS w/ note7c02…
20:31:19Switchboardscope.checkRun kept in scope · no expansione4d8…
20:29:44Indexevidence.hash4 sources sealed · 1 conflict flaggedb3aa…
20:12:03Ownergoal.createCrown weekly ops brief · budget 6 callsf190…

Receipts never rewritten · restore creates a recovery copy, never overwrites

Gates & policy

Fail-closed · a model opinion cannot approve work

Evidence & claims

CLEAR

Mechanical verification + human review. Source hashes run per workflow.

Information boundary

CLEAR

Outgoing payloads scanned for secret patterns. Heuristic, not comprehensive.

Execution authority

OWNER HOLD

Host-enforced call limits. Durable hold survives restart. Model replies cannot expand grants.

Publication

ABSENT

No publish/send tool installed. Export uses owner-selected native Save dialog.

Policy simulator

Before any run: shows what leaves the device, which tools may fire, what needs approval.

Drills

Make-it-fail: provider outage, bad evidence, injection attempt — next drill Fri 02:00.