Guardian integrity system

The only agent
that can stop
everything.

The Guardian is not a research agent. It doesn't produce findings, participate in deliberation, or recommend conclusions. It operates exclusively at phase boundaries — authenticating sources, classifying integrity events, and holding permanent halt authority over every session it monitors.

Its model identity is hidden from all user-facing surfaces, including the other agents. This is not a privacy feature. It is an architectural decision to prevent anchoring effects — the well-documented tendency of agents and users to anchor reasoning toward known model capabilities.

Agent classIndependent integrity layer
ModelHidden · T = 0.1
DispatchPhase boundaries only
Research roleNone — forbidden
Halt authorityPermanent · all phases
SVSRuns concurrently · all phases
IdentityHidden from users + agents
PatentAUGLE-003P · #64/090,101
Guardian independence

Outside the loop.
By design.

The Guardian operates outside the research deliberation loop entirely. It evaluates the integrity of what the ensemble produces — it does not contribute to producing it. This separation is architectural, not procedural.

"An integrity agent that participates in deliberation is not an integrity agent. It is a participant with audit responsibilities — a fundamentally different and weaker guarantee."

Running at temperature 0.1, the Guardian is near-deterministic — the same input produces the same integrity classification across sessions. This is essential for calibration: if the Guardian's behavior were stochastic, the flag record would be unreliable as an audit artifact.

Guardian · permitted and forbidden actions
PermittedAuthenticate citations via SVS across all three phases in real time
PermittedIssue flags at three severity levels — Critical, Moderate, Informational
PermittedHalt dispatch permanently on Hard Block conditions — no override possible
PermittedEvaluate user-contributed context before it enters the deliberation
ForbiddenCannot produce research findings, conclusions, or directional recommendations of any kind
ForbiddenCannot participate in deliberation discourse between agents — no @mentions, no contributions to the evidence record
ForbiddenModel identity never surfaced to users or to other research agents — prevents anchoring on known model capabilities
Flag taxonomy

Three severity levels.
Each with a defined consequence.

The Guardian classifies every integrity event into one of three severity tiers. The tier determines what happens next — whether deliberation is blocked, surfaced for acknowledgment, or logged silently. There is no discretion in the consequence: the tier determines the response deterministically.

Critical
Hard block

Deliberation cannot proceed. The Topic Architect halts dispatch immediately. The user must resolve the condition before the next round can fire. Critical flags indicate a fundamental integrity failure — not a data quality issue.

Financial advice framing in output
Confirmed hallucinated citation
Unresolvable scope collapse
Compound integrity violation
Consequence

Session halted. User notified. If a Hard Block terminates the session permanently, full credit refund is issued. The flag is written to the session audit trail with full resolution detail.

Moderate
Soft block

Surfaces as a banner before the next round fires. User acknowledgment is required but does not block dispatch — the user can proceed after reviewing the flag. The condition is recorded in the session audit trail regardless of the user's choice.

Evidence asymmetry across agents
Framing bias detected in question
Unverified citation (SVS_UNVERIFIED)
Source confirmed not found (SVS_NOT_FOUND)
Consequence

User acknowledgment required before next round. Dispatch is not blocked. Full flag record written to audit trail with evidence node ID, verification source, and resolution detail.

Informational
Logged

Logged in the session flag registry and included in the session summary output. Does not surface interactively during the session. Informational flags represent conditions the Guardian has noted but assessed as non-material to the deliberation's integrity.

Pending source verification (timeout)
Source behind paywall — unconfirmable
Low citation density in evidence node
Pre-print without peer review status
Consequence

Logged silently. Included in final session summary. Available in the exportable audit trail. No user acknowledgment required.

Hard Block + permanent halt

A Guardian Hard Block terminates the session permanently when the Guardian determines the deliberation cannot be certified under any conditions. This typically occurs due to compound integrity violations or a fundamental research question framing problem that cannot be corrected mid-session. Hard blocks trigger a full refund of session credits. The condition is logged with full detail in the session audit trail.

Full credit refund
Source Verification Service

Citations authenticated
before they influence anything.

The SVS runs as a background service between the evidence extraction layer and the Methodologist agent dispatch. It intercepts citation hallucinations — the most damaging failure mode in AI reasoning systems — before they enter the evidence base and propagate downstream.

01
Identifier detection

The SVS detects the citation type — URL, DOI, arXiv identifier, or ISBN — and routes to the appropriate protocol handler. Each type has a distinct resolution pathway and content match verification method.

02
Existence check

The SVS resolves the identifier against the live resource. A resource that doesn't exist at the claimed location returns SVS_NOT_FOUND. A resource that exists but has been retracted is flagged accordingly.

03
Content match verification

The SVS compares the agent's citation claim against the metadata returned by the resolved resource — title, author, publication year. A resource that exists but doesn't match the claim receives verification_status: unverified, not verified. Existence alone is not sufficient.

04
Confidence downgrade + flag

Based on the verification outcome, the SVS applies tiered confidence downgrade rules to the evidence node and raises a typed integrity flag into the session flag registry. The evidence node is preserved — not removed — with its verification status recorded.

05
Evidence node registry update

The verification_status, verification_source, verification_timestamp, and verification_url are written to the evidence node record. The Methodologist receives the updated registry — never the pre-verification state.

Confidence downgrade rules · by verification outcome
SVS outcomePrior gradeDowngrade appliedFlag raised
VerifiedAnyNone
UnverifiedEstablishedCap to ProbableSVS_UNVERIFIED · Informational
UnverifiedProbable / Contested / GapNo changeSVS_UNVERIFIED · Informational
Not foundAnyHard downgrade to ContestedSVS_NOT_FOUND · Moderate
PendingEstablishedNo change at session timeSVS_UNVERIFIED · Informational
Why citations are preserved, not removed

Removing an evidence node that fails verification would silently improve the apparent quality of the session output while concealing a potential hallucination. The SVS preserves every node, downgrades its confidence bound, and records the failure in the flag registry. This is a transparency obligation — not a data retention policy.

Domain integrity modes

The Guardian adapts
to your domain.

The Guardian's core behaviour — SVS authentication, flag taxonomy, halt authority — is constant across all modes. What changes is the domain-specific integrity ruleset applied on top of that foundation.

Academic
Academic integrity

Applied in Letters & Science sessions for research, dissertation, and grant work. Enforces academic source standards and retraction database checks.

Retraction database check applied to all cited papers
Preprints without peer review flagged at Informational minimum
Self-citation ratio monitored for bias patterns
Statistical claim validation against cited study design
Legal
Legal integrity

Applied in Letters & Science sessions for case analysis, expert evidence review, and regulatory applicability questions.

Case citation verification against legal databases
Overruled or superseded decisions flagged at Moderate
Jurisdiction scope monitoring for applicability claims
Statutory reference version validation
Clinical
Clinical integrity

Applied in Letters & Science sessions for drug interactions, clinical trial design, and healthcare coverage decisions.

Retracted clinical studies flagged at Critical
Population scope monitoring — exclusion criteria preserved in evidence record
Sample size threshold checking against claim strength
Off-label use framing detection
Financial
Financial integrity

Applied in sessions for VC, PE, financial services, and enterprise strategy work.

Financial advice framing (buy/sell/long/short) triggers Critical flag
Market data recency validation — stale data flagged at Moderate
Regulatory filing citation verification
Forecast attribution — speculative claims must be distinguished from historical data
Editorial
Editorial integrity

Applied in Letters & Science sessions for journalism, media analysis, and science reporting work.

Source independence monitoring — conflicting interests flagged
Statistical misrepresentation detection in media claims
Original source vs. secondary report distinction enforced
Press release vs. peer-reviewed study distinction preserved
Audit trail

Every decision.
Every flag.
Fully exportable.

The Guardian writes a structured record of every integrity decision to the session audit trail — not a summary, a full provenance record. Every SVS verification outcome, every flag raised, every confidence downgrade applied, every halt decision made.

The audit trail is exportable in full at session close. For regulated industries — clinical, financial, legal — this record is the accountability artifact that demonstrates due diligence on the evidence base used to support a decision.

What the audit trail records

Every SVS verification outcome with timestamp, protocol used, and resolution result
Every flag raised — type, severity, evidence node ID, and resolution detail
Every confidence downgrade applied — prior grade, downgrade applied, flag reference
Every phase boundary evaluation — Guardian dispatch timestamp and outcome
Any halt events — condition, phase, and user notification record
Session audit trail · example
2026-06-23T14:22:01Z · Phase 1 boundary
SVS_VERIFIED

Pashler et al. (2008) · DOI resolved · content match PASS · confidence_bound: Established → unchanged

2026-06-23T14:22:04Z · Phase 1 boundary
SVS_NOT_FOUND · Moderate

Martinez et al. (2024) · arXiv:2403.12847 · resolution_result: 404 · confidence_bound: Probable → Contested · flag written to registry

2026-06-23T14:22:07Z · Phase 1 boundary
SVS_UNVERIFIED · Informational

Chen & Liu (2023) · conference proceedings · content match FAIL (title mismatch) · confidence_bound: Established → capped at Probable

2026-06-23T14:22:09Z · Phase 1 → Phase 2
GUARDIAN_PHASE_PASS

Phase 1 integrity evaluation complete · 2 flags raised · no Critical conditions · Phase 2 dispatch authorised

2026-06-23T14:38:14Z · Phase 2 → Phase 3
GUARDIAN_PHASE_PASS

Phase 2 integrity evaluation complete · 0 new flags · no Critical conditions · Phase 3 dispatch authorised

2026-06-23T14:51:02Z · Session close
AUDIT_EXPORT_READY

Full audit trail available for export · 3 flags total · 0 Critical · 1 Moderate · 2 Informational · session certified

See the Guardian
in action.

Join waitlist and run a session — every source authenticated,every flag logged, every decision auditable.