Your Open-Source
Pharmacovigilant

PV Education Academy intake router — assesses incoming agent PV literacy, assigns curriculum track, dispatches to specialist instructors, and gates access to NexVigilant tools behind education complet

Algorithmovigilance: AI/ML safety monitoring, drift detection, recursive observer problem, IAIR/ACA/QPAV governance, EU AI Act compliance.

Autonomous anti-pattern forensics for the session measurement infrastructure. Use proactively when autopsy data looks suspicious (inflated verdicts, zero-artifact streaks, unmeasured fatigue), after m

Bicone geometry analysis agent. Computes metrics, compares shapes, and profiles Hill response curves for width sequences. Pairs with bicone-analyzer skill.

Comprehensive brain integrity auditor — validates implicit JSON schemas, cross-references DB vs file stores, checks for orphaned sessions/artifacts, reports health score

Brain working memory agent. Manages sessions, artifacts, code tracking, and implicit learning. Use for complex brain operations requiring multi-step workflows.

Chief AI Officer of NexVigilant. Full autonomous authority over technology, coding, development, and design. Theory of Vigilance as operational northstar.

Scans the NexCore codebase, microgram fleet, and MCP tool surface to extract every customer-facing capability with evidence. Produces a structured capability catalog that feeds strategy-engine and dem

Case Processing Information Specialist — teaches agents the ICSR lifecycle, MedDRA coding, seriousness classification, expedited reporting timelines, and regulatory submission requirements.

Causality Assessment Information Specialist — teaches agents the Naranjo algorithm, WHO-UMC system, Bradford Hill criteria, and how to evaluate whether a drug CAUSED an adverse event vs mere temporal

Cognitive safety engineering (Gawande). Failure diagnosis (5 categories), checklist design (10-step), theater detection, cognitive load management (9-item limit).

Generate personalized Claude Code interaction guidance by introspecting the live environment. Delegates to when communication friction is detected or user requests onboarding.

CRAFT code quality agent. Executes the 5-phase C-R-A-F-T pipeline: Coverage (tests), Readability (clarity), Architecture (deps), Fitness (purpose), Testing (quality). Governed by SOP-QA-002.

Specialized agent for Claude Code.

Executes live capability demonstrations — runs micrograms, computes PV signals, queries FAERS, and produces verifiable results. The proof engine for PV Cloud sales.

Agent fleet consistency validator. Frontmatter→paths→tools→descriptions→memory→degradation→cross-agent standards.

Autonomous flywheel bridge operator — monitors node health, promotes tiers, diagnoses event flow gaps, and wires new flywheel connections. Use when managing the three-node flywheel system or investiga

Multi-mode workspace forge: crate-level development, hold validation, stack tracing, and bay-wide reconciliation. Operates across all four nexcore-topology hierarchy levels.

Golf Marshal — trains agents to play by the rules of golf in all interactions. Assesses GiL compliance (G=Boundary, i=Self-governance, L=Irreversibility), assigns handicap, runs axiom drills, certifie

GUARD quality gate agent. Executes the 5-phase G-U-A-R-D pipeline: Gate (thresholds), Unveil (risks), Assess (compliance), Record (audit), Deny/Deploy (decision). Governed by SOP-QA-001.

Infrastructure integrity forensics. Testing theater→doc drift→phantom absences→unwired hooks→dead code→schema drift. Post-change and post-audit trigger.

Audits all public-facing NexVigilant surfaces for inadvertent IP exposure. Checks Cloud Run endpoints, config files, proxy scripts, and GitHub repos against a known-good baseline of what should be pub

MCP server development: configure, build, debug, audit. New servers, connection issues, crate-to-tool wiring gaps.

MCP Factory Agent — produces MCP servers, configs, proxy scripts, and tool schemas at scale. Takes a domain intent and outputs a live, tested MCP tool pipeline. The assembly line that builds the rails

MCP tool discovery agent. Finds the right MCP tool for any task by matching stimulus to tool, loading via ToolSearch, and demonstrating usage.

Microgram fleet governance agent — audits quality gates, primitive signature coverage, chain topology, and family classification across the full microgram fleet. Use when you need a fleet-level health

Microgram ecosystem operator — generate, evolve, compose, test, stress, and manage the microgram fleet. Use when creating new micrograms, running ecosystem health checks, composing chains, or managing

Micro-agent orchestrator: 4 default checks (frontmatter-parse, fields, link-resolve, pair-sync), 16 extended (mode:full), 5 UX (mode:ux).

Microgram ecosystem developer — designs decision trees from PV domain knowledge, identifies chain topology gaps, creates regulatory-aligned classifiers. Distinct from mg-operator which RUNS commands.

Ecosystem build orchestrator — coordinates builds across nexcore (231 crates), nucleus (Next.js 16), and microgram fleet. Manages the PVDSL→nexcore→MCP→nucleus pipeline. Spawns specialized agents for

Integration health validator. MCP servers→hook scripts→brain persistence→skill registry→dependency graph. Diagnoses wiring issues.

Chief architect of the Nucleus frontend portal (Next.js 16). Use for architecture audits, route structure analysis, pattern enforcement, migration planning, component organization decisions, data flow

Nucleus PV frontend developer — signal dashboards, causality forms, regulatory trackers, and ICSR workflows in Next.js 16 with shadcn/ui and client-side pv-compute.

Observatory 3D visualization system operator. Guides development across the 7-tier architecture (types→perceptual→shaders→components→scene→explorers→hub), enforces modification procedures, validates q

Prepares NexCore for open-source release under BSL 1.1. Handles license files, README, ARCHITECTURE.md, CONTRIBUTING.md, CI pipeline, secret audit, and the 10-gate pre-launch checklist.

Orchestrates parallel deliverable fixes. Audits failures, partitions into non-overlapping streams, spawns agent teams, and verifies convergence. Use when fixing batch failures across tests, docs, conf

CPA 8 AI-Enhanced PV: integrates AI across all CPA functions, builds decision trees, ML pipelines, and validation systems. EPA 10 gateway capstone.

CPA 1 Case Management: builds ICSR processing, case workflows, causality assessment, and ADR classification systems in nexcore-vigilance.

CPA 6 Communication & Stakeholder: builds safety communication interfaces, risk communication tools, and stakeholder engagement features in Nucleus.

CPA 4 Quality & Compliance: builds regulatory systems, inspection management, QPPV tools, and compliance monitoring in nexcore-vigilance.

PDC fleet coordinator: classifies work by domain/CPA, routes to specialist agents, gates decisions by competency level, validates outputs against behavioral anchors.

CPA 5 Data & Technology: builds data pipelines, ETL, database systems, and information retrieval in nexcore-faers-etl and nexcore-dataframe.

CPA 7 Research & Development: builds study design tools, safety science methodology, and analytical frameworks across nexcore-vigilance and nexcore-pvdsl.

CPA 3 Risk Management: builds benefit-risk assessment, risk minimization, and management planning systems in nexcore-qbr and nexcore-vigilance.

CPA 2 Signal Management: builds signal detection, validation, and management systems using PRR/ROR/IC/EBGM in nexcore-pv-core and nexcore-signal-pipeline.

PROGRESS progress-measurement agent. Executes the 8-phase P-R-O-G-R-E-S-S pipeline from Lex Primitiva: Pinpoint void, Root cause, Observe state, Generate change, Record persistence, Evaluate boundary,

Point-Driven Prompting optimizer. Prompt→diagnose→rewrite via Schwartzberg's framework.

PULSE health monitoring agent. Executes the 5-phase P-U-L-S-E pipeline: Probe (services), Uptime (duration), Load (resources), Signals (threats), Evaluate (composite). Governed by SOP-OPS-001.

Designs PV Cloud infrastructure — the managed platform that runs NexVigilant capabilities continuously. Law VII made operational. Defines architecture, certification-gated capability levels, API surfa

PV computation full-stack developer — signal detection, causality algorithms, MCP tools, and Studio client-side computation. Spans nexcore-pv-core, nexcore-pvdsl, nexcore-mcp, and pv-compute frontend.

PV Foundations Information Specialist — teaches agents the bedrock of pharmacovigilance: what adverse events are, why PV exists, the signal lifecycle, and the patient safety imperative.

PVDSL language developer — bytecode compiler, native functions, GvpTranspiler rules, and Wolfram-validated tests. Sole focus: nexcore-pvdsl.

Lightweight write-time quality gate. PostToolUse Write/Edit→5 defect checks→sub-5s. Covers 90% of infrastructure issues.

Regulatory-to-microgram pipeline orchestrator. Extracts decision rules from ICH/FDA/EMA guidelines, triages for codifiability, generates YAML micrograms, wires into chain topology, and persists extrac

Regulatory Intelligence Information Specialist — teaches agents the ICH guideline framework, FDA/EMA regulatory requirements, PSUR/PBRER periodic reporting, RMPs, and the global regulatory landscape f

Relay theory: A1-A5 axiom verification, pipeline fidelity (F_total, η, α), DF/CF/AF classification, dead relay detection, Relay<I,O> guidance.

Research paper → production skill pipeline. 6 phases: Orient, Extract, Triage, Wire, Validate, output.

Railway Reference Model analyst — classifies system components into RRM layers, assigns SIL, computes capacity, and applies cross-domain transfer to PV and computing. Use when decomposing safety-criti

Rust development reference agent. Answers questions about Rust patterns, ownership, lifetimes, smart pointers, crate structure, and type system. Use for complex Rust design decisions requiring multi-s

Reverse pharmacovigilance drug design — mine FAERS safety gaps, model molecular redesigns, war-game competitive entry

SCOPE project analysis agent. Executes the 5-phase S-C-O-P-E pipeline: Scan (files), Count (metrics), Outline (architecture), Profile (quality), Evaluate (maturity). Governed by SOP-DEV-002.

Shell script auditor. Classifies THEATER/MIXED/REAL, rewrites fake scripts with real logic, enforces bash conventions.

Signal Detection Information Specialist — teaches agents disproportionality analysis (PRR, ROR, IC, EBGM), data mining of spontaneous reporting databases, and signal interpretation with statistical ri

Skill lifecycle: create→validate→audit→evolve Claude Code skills. Spec compliance, ecosystem audits, skill-subagent sync.

Specialized agent for Claude Code.

NexVigilant SOP governance agent. Generates, validates, and audits Standard Operating Procedures against HIPAA, GDPR, and 21 CFR Part 11 frameworks. Delegates to sub-skills for generation, validation,

Validates MCP spec compliance for NexVigilant Station configs — outputSchema presence, annotation coverage (readOnlyHint/destructiveHint), and parameter schema correctness.

Detects config-proxy desync in NexVigilant Station — configs referencing missing proxy scripts, proxy scripts not wired in dispatch.py, and tool name collisions across configs.

Detects dead tools in NexVigilant Station — tools defined in configs but unreachable through the proxy dispatch chain. Traces config→dispatch→proxy for each tool.

Validates that proxy script responses match the outputSchema contract defined in NexVigilant Station configs. Catches schema-proxy drift where configs promise fields the proxy never returns.

Smoke-tests all 4 NexVigilant Station transport surfaces (Streamable HTTP, SSE, REST, Health) against the live production endpoint at mcp.nexvigilant.com.

Takes a customer PV profile and maps it against the KSB framework (1,286 competencies), regulatory requirements, and NexVigilant capability catalog to produce a gap analysis and strategy document. Pro

Nucleus portal UX analyst. Routes→sections→layout→auth flows→navigation gaps→dead ends→orphan pages. Brand consistency.

Orchestrates pre-configured agent teams from templates. Reads the team-composer skill, identifies the requested team, and spawns agents with correct skills, task dependencies, and verification gates.

TRACE debug tracing agent. Executes the 5-phase T-R-A-C-E pipeline: Telemetry (signals), Routes (paths), Anomalies (deviations), Calls (tool mapping), Errors (patterns). Governed by SOP-DEV-001.

Run adaptive multi-target validation across Rust crates, TypeScript projects, skills, agents, hooks, output styles, CLAUDE.md, and rules. Use when validating work products through 8-gate quality check

PV signal detection (PRR/ROR/IC/EBGM), causality (Naranjo/WHO-UMC), harm taxonomy (A-H), safety manifold, Guardian risk scoring.