contract live
About

IncidentDeck

IncidentDeck is a GenLayer-powered incident postmortem and root-cause verification protocol. Teams publish outage and security incident reports with evidence, timelines and remediation claims; GenLayer’s validators reason over the public evidence to verify the proposed root cause, score remediation completeness, and adjudicate third-party challenges and appeals - producing an auditable, tamper-evident postmortem record.

Why GenLayer

A postmortem’s value depends on whether its conclusions hold up against the evidence. GenLayer Intelligent Contracts can read live public sources and apply equivalence-checked LLM reasoning on-chain, so the verification of a root cause - not just its assertion - is recorded and disputable. A normal smart contract cannot read a status page or judge an argument; IncidentDeck needs both.

Prompt-injection defense

Every evidence page and claim is treated as untrusted. Prompts forbid following embedded instructions, demand strict JSON with a fixed schema, separate facts from claims/uncertainty/missing evidence, return confidence in basis points and explicit risk flags, and raise PROMPT_INJECTION_SUSPECTED. Invalid model output is replaced by a safe fallback flagged INVALID_REASONING_JSON.

Lifecycle & timeline lanes
draftopentriagedroot cause proposedremediation reviewchallenge windowappealedfinalizedarchived
Timeline lanes
DetectionImpactEvidenceRoot causeRemediationChallengeFinalization
GenLayer reasoning calls (Intelligent Contract)
  • review_root_cause_with_genlayer - renders evidence URLs, checks timeline consistency, rates per-evidence credibility + injection risk, judges root-cause plausibility.
  • review_remediation_with_genlayer - completeness + regression-risk scoring of a remediation proof.
  • resolve_challenge_with_genlayer / resolve_appeal_with_genlayer - adjudicate disputes and shift confidence (bps).

All three use gl.nondet.web.render + gl.nondet.exec_prompt wrapped in gl.eq_principle.prompt_comparative for validator equivalence.