S271
RAG Evidence Coverage Auditor
Audit whether outputs are evidence-backed; list unsupported claims; propose retrieval queries to close gaps.
Inputs:Strategic Priority, Reference Sources
Output:Coverage audit + gap list + query plan.
30 Services
AI Assurance
AI governance, retrieval-augmented generation quality, audit trails, and compliance frameworks.
All Services
30 services in this domainS272
Contradiction Resolver (Cross-Source)
Identify contradictions; resolve by evidence weighting and recency; if unresolved, mark UNKNOWN with next steps.
Inputs:Deadline, Reference Sources
Output:Contradiction log + resolution rationale + verification tasks.
S273
"No-Source, No-Claim" Enforcement Protocol
Rewrite outputs to enforce strict citation; remove or label any unsupported assertions.
Inputs:Strategic Priority
Output:Cleaned output + evidence map + residual unknowns.
S274
Hallucination Risk Stress Test (Adversarial QA)
Generate adversarial questions to probe weak points; require evidence-backed answers; flag failures.
Inputs:Strategic Priority
Output:Red-team Q set + failure log + mitigations.
S275
Input Injection Defense Hardening (RAG)
Define rules to treat docs as data; detect malicious instructions; sanitize retrieval; log injections.
Inputs:Reference Sources
Output:Defense protocol + detection checklist + audit steps.
S276
Provenance Ledger Schema (Evidence and Versioning)
Define provenance fields: source IDs, timestamps, claim links, version hashes.
Inputs:Strategic Priority
Output:Ledger schema + usage SOP + audit checklist.
S277
Evaluation Harness Design (Quality + Safety)
Specify evaluation tests: factuality, coherence, bias, compliance, reproducibility; define pass/fail thresholds.
Inputs:Strategic Priority
Output:Test suite spec + scoring rubrics + governance.
S278
Bias and Fairness Audit (Policy and Outputs)
Audit for biased framing; propose neutral rewrites; ensure sovereignty-respecting language.
Inputs:Strategic Priority
Output:Bias audit + rewrites + compliance notes.
S279
Model Risk Management Policy (Embassy/Institute)
Draft MRM policy: acceptable use, review gates, incident response, documentation, retraining cadence.
Inputs:Strategic Priority
Output:Policy + roles + templates.
S280
Secure Instruction Library Governance (Enterprise)
Define instruction library governance: approvals, versioning, access control, change logs.
Inputs:Strategic Priority
Output:Governance SOP + templates + audit checklist.
S281
Data Minimization and Retention Controls for RAG
Define minimization rules, retention, access logging, and export gating by sensitivity.
Inputs:Strategic Priority
Output:Controls policy + implementation checklist.
S282
“UNKNOWN Management” Protocol (Decision Safety)
Define how UNKNOWNs are tracked, verified, and resolved; include decision trigger linkage.
Inputs:Strategic Priority
Output:UNKNOWN protocol + tracker template + escalation gates.
S283
Reproducibility Packet Generator (Run Metadata)
Generate reproducibility metadata: inputs, instruction versions, retrieval scope, assumptions, output hashes.
Inputs:Strategic Priority
Output:Reproducibility packet + audit appendix.
S284
Evidence Weighting Rubric (Recency/Authority/Independence)
Define evidence weighting rubric; apply to retrieved sources; document conflicts.
Inputs:Strategic Priority
Output:Rubric + applied evidence weights + contradiction notes.
S285
Multi-Language Source Diversity Planner
Plan multilingual retrieval (Khmer/Thai/Vietnamese/Chinese/English where available); define translation QA.
Inputs:Strategic Priority
Output:Diversity plan + QA checklist + audit steps.
S286
RAG Query Expansion Generator (Cambodia/ASEAN)
Generate 20–40 targeted queries by topic; include synonyms and institutional terminology.
Inputs:Strategic Priority
Output:Query pack + retrieval plan + evidence targets.
S287
Structured Output Schema Validator (Briefs/Clauses/Maps)
Validate output conforms to schema; list missing sections; regenerate only missing parts with evidence.
Inputs:Strategic Priority
Output:Validation report + patched sections + audit appendix.
S288
Safety Gatekeeper Protocol (No Illicit Assistance)
Check requests for illicit content; refuse unsafe parts; provide safe alternatives; log rationale.
Inputs:Strategic Priority
Output:Safety decision + safe alternatives + compliance note.
S289
Confidence Calibration Protocol (Per Section)
Assign confidence per section tied to evidence density; explain what would raise confidence.
Inputs:Strategic Priority
Output:Confidence map + evidence gaps + next steps.
S290
Scenario Branch Consistency Checker
Ensure scenario branches don’t contradict baseline facts; reconcile; label uncertainties.
Inputs:Strategic Priority
Output:Consistency report + corrected branches + audit log.
S291
Decision-Assist Boundary Setter (No Proxy Overreach)
Ensure outputs present options and triggers; avoid unauthorized “decide for user”; enforce human gates.
Inputs:Strategic Priority
Output:Boundary-compliant outputs + decision gates + warnings.
S292
Sensitive Export Gate (Official-Sensitive Mode)
Add export checklist: approvals, redactions, attribution checks, evidence verification.
Inputs:Strategic Priority
Output:Export gate checklist + redaction guidance.
S293
Quality Rubric (10/10 Audit Standards)
Define rubric: accuracy, provenance, clarity, feasibility, compliance, risk controls; score outputs and justify.
Inputs:Strategic Priority
Output:Scored rubric + improvement actions + re-run plan.
S294
“Claim-to-Evidence Map” Generator
Link each claim to evidence; list orphan claims; propose retrieval or removals.
Inputs:Deadline, Reference Sources
Output:Claim-evidence map + orphan list + remediation.
S295
Version Diff Explainer (Run v1 vs v2)
Compare versions; explain changes, drivers, evidence updates, and impacts on recommendation.
Inputs:Deadline
Output:Diff narrative + change log + updated risks.
S296
Audit-Ready Executive Summary Composer
Produce a concise exec summary that references evidence and clearly states uncertainties.
Inputs:Strategic Priority
Output:200–400 word summary + evidence anchors + unknowns.
S297
Service Pack Builder (Scenario-Specific Set of 12)
Select 12 best service configurations from this catalog for the scenario; order them for workflow; explain why.
Inputs:Strategic Priority
Output:12-instruction workflow + run order + expected artifacts.
S298
Secure Automation Hook Spec (Queue + Webhooks)
Define automation hooks: run triggers, audit logging, approvals, export; include security controls.
Inputs:Strategic Priority
Output:Automation spec + job types + governance.
S299
Knowledge Base Update Proposal (From New Evidence)
Propose KB updates: new tags, deprecations, contradiction merges, and governance approvals.
Inputs:Deadline, Reference Sources
Output:Update proposal + change impacts + audit steps.
S300
Full Helios “Deep Scan” Run Template (Reusable)
Execute deep scan: retrieval plan, contradiction resolution, scenario testing, compliance gates, and audit appendix; fail closed on missing evidence.
Inputs:Strategic Priority, Reference Sources
Output:Deep scan report + reproducibility packet + audit-ready artifacts.