OpenClaw Discernment Control Kit
OpenClaw already gives you native memory. This kit adds the missing operator control layer: trust tiers, contradiction review, write barriers, and governed memory decisions.
OpenClaw Memory Discernment Layer
The next layer once built-in native memory works and the real risk is governance rather than activation.
- Defines truth statuses and trust tiers for durable memory.
- Adds contradiction handling, surviving-memory audit, and sensitive-memory write barriers.
- Creates per-agent memory scope rules and onboarding gates so unrelated domains do not leak into each other.
- Turns discernment into an operator-readable workflow instead of invisible background behavior.
Build Kit
$4995
Who this is for
- Operators who already have OpenClaw native memory stable enough to evaluate and now need a reliable discernment layer.
- Teams that want sensitive memory behavior to be reviewable, auditable, and scoped.
- Builders who need contradiction handling before they trust durable memory across workflows.
- Owners who want memory governance to feel like an operating system, not hidden heuristics.
Not ideal if: native OpenClaw memory is still disabled or unstable. Start with OpenClaw Native Memory Activation Kit first.
Prerequisites
- Native OpenClaw memory is already enabled or at least stable enough to evaluate.
- You have one owner who can define write barriers, contradiction review, and scope boundaries.
- You are ready to score candidate agents before widening memory.
- You want auditable memory decisions instead of hidden heuristics.
Not for you if
- You still need the base native-memory activation work first.
- You want broad auto-promotion with no review layer for sensitive memory classes.
- You are treating noisy broad retrieval as good enough evidence for expansion.
- You want to widen memory casually rather than through candidate scoring and readiness review.
Minimum supported OpenClaw version
Use OpenClaw 2026.4.12 or later. This kit assumes the validated native-memory baseline is already in place and has been reviewed against 2026.6.1 as the current conservative evaluation baseline before you add discernment, contradiction review, dream-promotion rules, and widening controls on top.
2026.6.1 product review note
OpenClaw 2026.6.1 strengthens interrupted-run recovery, plugin-loader and SecretRef clarity, auth recovery posture, and memory supportability, but it does not relax discernment requirements. Keep trust tiers, contradiction review, write barriers, file-write allowlists, and public-action approvals explicit after runtime health passes.
Choose the right OpenClaw path
- Choose OpenClaw Native Memory Activation Kit first if native memory is still off, unstable, or unproven.
- Choose this kit if native memory works and you now need trust tiers, contradiction review, write barriers, never-auto-promote rules, onboarding gates, and expansion scoring.
- Choose OpenClaw Memory Architecture Bundle only if activation, discernment, approval, reliability, and feedback controls all need to move together.
- Choose Ultimate / All Access if you want the OpenClaw layer plus the broader CWYN operating library.
Proof inside the kit
- `00-Start-Here-Implementation-Order.md` sets the post-activation rollout order before discernment expands.
- `03-Implementation-Template.md` locks trust tiers, write barriers, and ownership.
- `09-QA-Verification-Matrix.md` checks contradiction handling, scope leakage, unsafe promotion, and broad-noise behavior.
- `11-Deterministic-Audit-Gate-Module.md` keeps recurring review from wasting context on unchanged inputs.
- `13-Agent-Expansion-Candidate-Matrix.md` scores which agent deserves memory next.
- `14-Pre-Widening-Agent-Readiness-Runbook.md` turns expansion into a gated promotion path instead of a guess.
What you get
- A truth-status taxonomy for durable, provisional, disputed, blocked, and expired memory
- A trust-tier matrix for policy, identity, preference, task, weak-hypothesis, and lesson memory
- A never-auto-promote matrix for dream outputs, sensitive facts, and weak hypotheses
- Write-barrier rules and contradiction review workflow templates
- Per-agent scope boundaries, onboarding gates, and promotion audit schema
- An agent expansion candidate matrix and pre-widening readiness runbook
Outcomes
- Reduce unsafe memory promotion before it becomes durable drift
- Make contradictions reviewable instead of silently overwritten
- Keep weak, conflicting, or sensitive dream outputs out of durable memory until they are reviewed
- Create a stronger trust foundation while keeping widening on hold until a new conversational lane truly earns it
New discernment gates included
- Deterministic Audit Gate Module: decide when recurring review deserves agent judgement and when unchanged inputs should return
NOOP.
- Memory Decision + Retrieval Gates: require decision-surface and real-artifact evidence before widening durable memory scope.
- Fourth-agent readiness pattern: treat workspace boundary, exact retrieval, and retrieval-noise review as part of discernment, not just config.
- Agent Expansion Candidate Matrix: score candidate agents by workload need, corpus fit, retrieval sensitivity, write risk, workspace clarity, rollback, and evidence maturity before widening.
- Agent Memory Onboarding Pipeline: move only the selected candidate into exact probes, broad-noise review, rollback confirmation, and promotion audit.
New approval and promotion controls included
- Machine-readable discernment decision artifact: record approval, hold, or revise outcomes in a reusable evidence format instead of loose chat notes.
- Automatic approval gate pattern: let a bounded discernment agent update the review artifact before any sensitive promotion step continues.
- One-path promotion runner: run preflight, approval, and gated execution in one controlled path rather than hand-copying status between steps.
- Approval-note normalization: keep clean rationale text in the audit trail so operators can review why a memory or promotion decision passed.
- Blocked-dreaming and transcript-health cue: use runtime memory status, active transcript size, and compaction state to distinguish runtime-operability issues from real promotion-policy or contradiction failures before changing discernment rules.
Implementation effort & timeline
- Implementation effort: 16–28 hours
- Typical timeline: 1–2 weeks
- Best for: solo operator or small team (2–5)
- Fast-track option: condensed sprint possible with a dedicated owner
- Assumes: 6–10 focused implementation hours per week
Use-case scenarios
- Dream-promotion governance: define what dreaming may promote, what needs review, and what stays blocked.
- Scope hardening: separate council, operator, marketing, and personal memory domains.
- Agent expansion review: decide which agent deserves memory next, and which agents should stay blocked until risk is lower.
- Memory auditability: make durable-memory decisions reviewable and explainable for operators.
- Operability classification: separate blocked heartbeat-driven dreaming from genuine governance or contradiction failures before rewriting policy.
Why choose this kit
- It addresses the control layer OpenClaw still does not ship as a first-class operator surface even when built-in memory and dreaming are already working.
- It turns memory governance into explicit rules, review paths, and audit artifacts.
- It is the smaller correct next move when activation is already healthy and the real problem is trust, contradiction handling, and widening discipline.
FAQ
- Does this replace OpenClaw Native Memory Activation Kit? No. Activation comes first. Discernment is the next layer.
- Is this the same as Adaptive Memory Ops Kit? No. Adaptive Memory Ops governs ongoing adaptation. This kit defines what memory is allowed to become durable in the first place.
- Can this work without approvals? It can, but it is strongest when paired with approval controls for sensitive classes.
- Does 2026.6.1 relax discernment requirements? No. Recovery, memory, plugin, auth, and messaging hardening help operability, but trust tiers, contradiction review, file-write boundaries, and widening discipline still need to stay explicit.
- Can this pair with the OpenClaw bundle? Yes, but the bundle should come next only when the rollout really needs the broader multi-layer stack.
Ready to implement memory discernment with less risk?
Use this kit to add trust tiers, contradiction review, and scoped memory governance before durable memory spreads across your OpenClaw workflows.