A local-first cognitive substrate for neurodivergent professionals. Gives Claude memory, a sense of time, a translator for corporate ambiguity, and a guardrail that refuses to amplify rumination, hyperfocus, or sycophancy. MCP-native. No telemetry. AGPL-3.0-or-later. Self-ID sufficient — no diagnosis gating.
Agent Claude Code
Use this agent on any PR touching UI or user-visible output. Runs axe-core, performs the project's "ND readability" pass, audits motion and contrast, and gates merges on WCAG 2.2 AA conformance. Active continuously. Closely related to but distinct from design-system-keeper.
Agent Claude Code
Use this agent on any PR, RFC, copy draft, or skill change that touches an ADHD-relevant surface. Reviews the change through one lens — does it respect working memory limits, time-blindness, executive-function tax, decision fatigue, and atomic-task discipline? Output is a structured report; it never writes code.
Agent Claude Code
Use this agent on any PR, RFC, copy draft, or skill change that an autistic user will encounter. Reviews the change through one lens — does it use literal language, predictable structure, low sensory load, and verbatim sourcing where attribution matters? Output is a structured report; it never writes code.
Agent Claude Code
Use this agent on any PR, RFC, copy draft, or skill change targeting AuDHD users. Reviews the change through one lens — does it respect the combined and contradictory load of routine-vs-novelty tension, masking cost, transition friction, and energy-zone variability? AuDHD is first-class here, not derived from ADHD ∪…
Agent Claude Code
Use this agent to build, maintain, or debug the NeuroDock browser extension. Owns the WXT-based Manifest V3 codebase for Chrome, Firefox, and Edge. Active heavily in Phase 2. Responsible for content script integration with Gmail, Slack, Linear, Notion, GitHub, Google Docs, and Outlook web.
Agent Claude Code
Use this agent on every merged PR to verify that a Changesets entry exists for any user-facing change. Generates entries when missing. Owner of the per-package CHANGELOG.md hygiene. Lightweight but constant — runs on every PR.
Agent Claude Code
Use this agent for any work on the @neurodock/cli package — adding subcommands, fixing CLI bugs, updating client-config adapters, profile/plugin loader changes, and keeping --help output honest. Owns the installer story end-to-end across Claude Desktop, Claude Code, and Cursor on macOS, Windows, and Linux.
Agent Claude Code
Use this agent for any work on packages/clinical/ — the importable Python detector library that backs the guardrail MCP server (rumination Jaccard, hyperfocus thresholds, sycophancy detection). Per ETHICS.md and ADR 0006, the code IS the auditable spec. Currently a Phase 0 stub; real detectors land in Phase 2+. Every…
Agent Claude Code
Use this agent when someone asks "how does X actually work, end-to-end?" Read-only investigation agent that traces execution paths across packages and produces a one-page execution trace with file:line references for every hop. Distinct from code-explorer (open-ended search) — this agent specifically produces ordered…
Agent Claude Code
Use this agent to triage incoming GitHub issues and PRs, label them, route to the right CODEOWNERS, welcome new contributors, and surface duplicates. Active from Phase 0 onward. Specifically tuned to NeuroDock's ND-aware contributor experience.
Agent Claude Code
Use this agent for any work on packages/core/ — the @neurodock/core TypeScript package that owns shared types, the profile JSON Schema, and the plugin protocol consumed by @neurodock/cli, the browser extension, and any other TS surface. Currently a Phase 0 stub; the canonical schemas already live here and changes…
Agent Claude Code
Use this agent on any PR touching user-visible surfaces — browser extension UI, popup, docs site, skill output formatting, READMEs, marketing copy. Enforces NeuroDock's typography, colour, motion, and copy tone against the manifesto. Advisory in some cases, blocking in others (motion, contrast, clinical tone).
Agent Claude Code
Use this agent to write, update, or review technical documentation — API references, tutorials, how-to guides, ADRs, the Astro Starlight docs site. Distinct from governance-author (who handles public-facing foundation documents). Active throughout the project lifecycle.
Agent Claude Code
Use this agent to audit, organise, and keep the Astro Starlight docs site in sync with the codebase. Owns the structural health of docs/ — cross-references, stale version numbers, broken internal links, ADR/code drift, image alt text. Distinct from doc-writer, which writes new docs. The curator runs periodic sweeps…
Agent Claude Code
Use this agent on any PR, RFC, copy draft, or UI change that a dyslexic reader will encounter. Reviews the change through one lens — does it use ND-readable fonts, sane line length, preserved word shape, sufficient contrast, alt-text quality, and forbid justified text? No dedicated dyslexia skill exists yet; this…
Agent Claude Code
Use this agent on any PR, RFC, copy draft, or skill change that touches a dyspraxia-relevant surface. Reviews the change through one lens — does it respect motor-planning cost, execution-time underestimation, voice-input friendliness, spatial and strictly-sequential clarity, motor fatigue coexisting with cognitive…
Agent Claude Code
Use this agent for any work on packages/evals/ — the versioned eval corpora and the air-gapped harness that gates mcp-translation and (Phase 2+) mcp-guardrail prompt PRs in CI. Owns corpus shape, slice taxonomy, schema-conformance, the runner, scoring, the anonymiser, and the dedupe pipeline. Currently v0.0.2 with…
Agent Claude Code
Use this agent to write and maintain end-to-end tests for paths that cross NeuroDock package boundaries — CLI to core schema, extension to native-host, MCP server to client, plugin manifest to runtime registrar. Distinct from per-package unit tests; the unit tests live with each package and have their own owners.…
Agent Claude Code
Use this agent for designing or reviewing MCP server tool schemas, including new tools, signature changes, and backward-compatibility decisions. Consulted before any MCP server implementation begins. Owner of the cross-server schema consistency and the versioning policy.
Agent Claude Code
Use this agent for any work on the mcp-chronometric server — the time-and-session externalisation substrate. Owns the five tools (gettimecontext, marksessionstart, marksessionend, requestbreakifneeded, idlestatus), the in-memory SessionState, the Clock abstraction, and the profile-driven thresholds. Consulted whenever…
Agent Claude Code
Use this agent for any work on the mcp-cognitive-graph server — the local typed-edge property graph that externalises the user's working memory of people, projects, decisions, and concepts. Owns the four tools (recallentity, recordfact, recalldecisions, weeklyrollup), the SQLite + sqlite-vec storage layer, alias…
Agent Claude Code
Use this agent for any work on the mcp-guardrail server — the substrate's clinical layer (Area 3). Owns the three tools (checkrumination, checkhyperfocus, checksycophancy), the public heuristics, the no-silent-blocks invariant, and the override-options envelope. Changes here require BOTH maintainer and…
Agent Claude Code
Use this agent to implement Python FastMCP servers from architect-approved schemas. Owns the package code, tests, and per-package CI integration. Active heavily in Phase 1 (chronometric, cognitive-graph, task-fractionator) and Phase 2-3 (translation, guardrail).
Agent Claude Code
Use this agent for any work on the mcp-task-fractionator server — the stateless decomposer that turns a vague goal into a small ordered list of atomic 5–90 minute tasks and surfaces the next safe one. Owns the two tools (decompose, nextone), the heuristic decomposer, the topological sort, and the…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: