Threat Model
| Document Control |
Value |
| Scope |
Public threat summary for the Devnet MVP |
| Status |
Informational but aligned with current implementation |
This page is a public summary. It is meant to show what classes of failure matter, not to expose private operational detail.
1. Main threats and mitigations
| Threat |
Failure mode |
Current mitigation |
Remaining limitation |
| Admin key compromise |
Unauthorized config or treasury actions |
Restricted authority surfaces, public documentation of roles |
Still centralized in MVP |
| Oracle key compromise |
Malicious pulse accepted |
On-chain Ed25519 verification against configured pubkey |
Single-signer trust remains |
| Replay or duplicate execution |
Duplicate pulse, claim, refund, or reveal path |
State flags, round binding, timing windows, signature checks |
Depends on correct implementation discipline |
| No-reveal griefing |
Users commit then disappear |
Burn-on-no-reveal and deterministic deadlines |
UX penalty remains real for inattentive users |
| Misconfiguration |
Wrong oracle key, mint, or timing |
Deterministic config, scripted deployment flows, public parameter docs |
Human error is still possible |
| Operator outage |
Rounds stall or refund path is needed |
Timeout and refund path, inspection scripts, automation |
Current liveness still depends on operator health |
2. Public invariants worth checking
| Invariant |
Why it matters |
| No commit after deadline |
Protects fairness of the pulse target |
| No reveal before pulse or after reveal deadline |
Prevents timing abuse |
| Pulse accepted once only |
Prevents contradictory round state |
| Claim cannot succeed twice |
Protects supply and payout correctness |
| Sweep cannot pre-empt grace |
Preserves winner claim window |
| Refund requires pulse absence |
Prevents invalid dual-resolution paths |
3. MVP limitations that should be stated plainly
| Limitation |
Current status |
| Single-oracle dependency |
Present in MVP |
| Centralized admin surfaces |
Present in MVP |
| Operational liveness |
Strongly helped by the supervisor pipeline |
| Devnet assumptions |
Parameters and cadence are tuned for testing, not production economics |