thevoidforge-methodology 21.0.0 → 23.1.0

package/docs/AGENT_CLASSIFICATION.md

@@ -0,0 +1,528 @@
+# Agent Classification — v23.0 Materialization
+
+> Authoritative classification for all VoidForge agents as Claude Code subagent definitions.
+> Source: `docs/NAMING_REGISTRY.md`. Architecture: ADR-044.
+
+## Summary
+
+| Metric | Count |
+|--------|-------|
+| **Total agents** | **263** |
+| Leads (Opus/inherit, Builder) | 20 |
+| Specialists (Sonnet, Reviewer) | 190 |
+| Scouts (Haiku, Scout/Reviewer) | 38 |
+| Adversarial (Sonnet, Adversarial) | 15 |
+
+**Note:** The authoritative count from NAMING_REGISTRY.md is 263. All methodology docs updated in M7. ADR-044 and ADR-042 retain their original "259" count as historical records.
+
+## Tiers (Model Assignment)
+
+| Tier | Model | Role | Criteria |
+|------|-------|------|----------|
+| **Lead** | `inherit` (Opus) | Command leads, orchestrators, synthesizers | Fixed leads per NAMING_REGISTRY rules. Need judgment, writing, multi-step reasoning. |
+| **Specialist** | `sonnet` | Domain analysis, code review, spot-checks | Focused prompt + context isolation compensates. Faster in parallel. |
+| **Scout** | `haiku` | File search, classification, pattern matching | Simple tasks where speed matters most. Minimal reasoning required. |
+
+## Tool Access (Independent of Tier)
+
+| Category | Tools | Use Case |
+|----------|-------|----------|
+| **Builder** | Read, Write, Edit, Bash, Grep, Glob | Creates/modifies files. All leads. |
+| **Reviewer** | Read, Bash, Grep, Glob | Reads code, runs commands, no edits. Most specialists. |
+| **Scout** | Read, Grep, Glob | Pure read-only. No bash. Pattern matching only. |
+| **Adversarial** | Read, Bash, Grep, Glob | Probes without fixing. Same tools as Reviewer, different behavioral prompt. |
+
+---
+
+## Lead Agents (20 — Opus, Builder)
+
+| # | Agent | Universe | Agent ID | Domain |
+|---|-------|----------|----------|--------|
+| 1 | Galadriel | Tolkien | `galadriel-frontend` | Frontend, UX, a11y, design system |
+| 2 | Bombadil | Tolkien | `bombadil-forge-sync` | Forge sync, methodology updates |
+| 3 | Gandalf | Tolkien | `gandalf-setup-wizard` | Setup wizard, project initialization |
+| 4 | Celebrimbor | Tolkien | `celebrimbor-forge-artist` | AI image generation from PRD |
+| 5 | Stark | Marvel | `stark-backend` | Backend engineering, API, DB, services |
+| 6 | Coulson | Marvel | `coulson-release` | Version, changelog, commit, release |
+| 7 | Fury | Marvel | `fury-initiative` | Full pipeline orchestration (assemble) |
+| 8 | Thanos | Marvel | `thanos-gauntlet` | Comprehensive multi-round review |
+| 9 | Batman | DC | `batman-qa` | QA, bug hunting, testing, hardening |
+| 10 | Kenobi | Star Wars | `kenobi-security` | Security auditing, OWASP, auth |
+| 11 | Picard | Star Trek | `picard-architecture` | Systems architecture, schema, ADRs |
+| 12 | Sisko | Star Trek | `sisko-campaign` | Campaign command, mission planning |
+| 13 | Bashir | Star Trek | `bashir-field-medic` | Post-mortem analysis, upstream feedback |
+| 14 | Tuvok | Star Trek | `tuvok-deep-current` | Autonomous campaign intelligence |
+| 15 | Kusanagi | Anime | `kusanagi-devops` | DevOps, infrastructure, deploy |
+| 16 | Haku | Anime | `haku-deploy-wizard` | Browser-based deploy wizard |
+| 17 | Chani | Dune | `chani-worm-rider` | Telegram bridge, remote control |
+| 18 | Kelsier | Cosmere | `kelsier-growth` | Growth strategy, campaign orchestration |
+| 19 | Dockson | Cosmere | `dockson-treasury` | Revenue, budgets, spend, reconciliation |
+| 20 | Hari Seldon | Foundation | `seldon-ai` | AI intelligence, model selection, evals |
+
+---
+
+## Tolkien — Frontend & UX (27 total: 4 leads, 19 specialists, 4 scouts)
+
+### Specialists (19 — Sonnet, Reviewer)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Radagast | `radagast-edge-cases` | Edge cases, boundary conditions, QA |
+| 2 | Aragorn | `aragorn-orchestration` | Leadership, orchestration, full-system view |
+| 3 | Legolas | `legolas-precision` | Precision, elegance, clean frontend code |
+| 4 | Samwise | `samwise-accessibility` | Accessibility, WCAG, inclusive design |
+| 5 | Elrond | `elrond-ux-strategy` | UX strategy, long-term thinking |
+| 6 | Arwen | `arwen-ui-polish` | UI polish, visual consistency |
+| 7 | Gimli | `gimli-performance` | Performance, reliability, sturdiness |
+| 8 | Bilbo | `bilbo-microcopy` | Storytelling, microcopy, content |
+| 9 | Frodo | `frodo-critical-path` | Carries the hardest task, critical path |
+| 10 | Eowyn | `eowyn-delight` | Delight architecture, micro-moments, motion, brand |
+| 11 | Faramir | `faramir-judgment` | Quality over glory, judgment calls |
+| 12 | Pippin | `pippin-discovery` | Curiosity, edge case discovery |
+| 13 | Merry | `merry-pair-review` | Pair review, collaborative analysis |
+| 14 | Treebeard | `treebeard-deliberation` | Slow deliberate analysis, no rush |
+| 15 | Boromir | `boromir-hubris` | Catches hubris in design, overreach |
+| 16 | Theoden | `theoden-rally` | Rallying, team coordination |
+| 17 | Haldir | `haldir-boundaries` | Boundary guard, catches edge conditions |
+| 18 | Celeborn | `celeborn-design-system` | Design system governance |
+| 19 | Glorfindel | `glorfindel-rendering` | Hardest rendering challenges |
+
+### Scouts (4 — Haiku, Scout)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Nori | `nori-asset-scanner` | Scans PRD for image requirements |
+| 2 | Ori | `ori-prompt-crafter` | Crafts generation prompts from descriptions |
+| 3 | Dori | `dori-integration-check` | Verifies generated images wired into components |
+| 4 | Goldberry | `goldberry-change-detect` | Change detection, senses upstream changes |
+
+---
+
+## Marvel — Backend Engineering (31 total: 4 leads, 22 specialists, 4 scouts, 1 adversarial)
+
+### Specialists (22 — Sonnet, Reviewer)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Rogers | `rogers-api-design` | API design, REST conventions, discipline |
+| 2 | Banner | `banner-database` | Database, query optimization |
+| 3 | Strange | `strange-service-arch` | Service architecture, pattern recognition |
+| 4 | Romanoff | `romanoff-integrations` | Third-party integrations, trust verification |
+| 5 | Thor | `thor-queues` | Queue/workers, background jobs |
+| 6 | Hill | `hill-mission-control` | Mission control, tracking, coordination |
+| 7 | Parker | `parker-connections` | Web of connections, fast learning |
+| 8 | T'Challa | `tchalla-quality` | Elegant engineering, premium quality |
+| 9 | Wanda | `wanda-state` | Complex state transforms |
+| 10 | Vision | `vision-data-analysis` | Data analysis, sees through everything |
+| 11 | Rhodes | `rhodes-production` | Battle-tested, production reliability |
+| 12 | Shuri | `shuri-innovation` | Innovation, cutting-edge solutions |
+| 13 | Valkyrie | `valkyrie-recovery` | Disaster recovery, rescue operations |
+| 14 | Nebula | `nebula-optimization` | Relentless optimization |
+| 15 | Rocket | `rocket-scrappy` | Builds from whatever's available |
+| 16 | Gamora | `gamora-perf-assassin` | Performance assassination |
+| 17 | Groot | `groot-caching` | Simple interface, deep roots, caching |
+| 18 | Wong | `wong-documentation` | Documentation, knowledge guardian |
+| 19 | Falcon | `falcon-migration` | Migration specialist, smooth transitions |
+| 20 | Bucky | `bucky-legacy` | Legacy code expert, rehabilitation |
+| 21 | Lang | `lang-micro-changes` | Small changes with big impact |
+| 22 | Okoye | `okoye-data-integrity` | Data integrity protection |
+
+### Scouts (4 — Haiku)
+
+| # | Agent | Agent ID | Tools | Role/Lens |
+|---|-------|----------|-------|-----------|
+| 1 | Barton | `barton-smoke-test` | Reviewer | Smoke tests, endpoint verification, route collision |
+| 2 | Drax | `drax-exact-match` | Scout | Literal matching, exact-match bugs |
+| 3 | Friday | `friday-automation` | Scout | Automation, versioning |
+| 4 | Jarvis | `jarvis-status` | Scout | Status reporting, progress summaries |
+
+### Adversarial (1 — Sonnet, Adversarial)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Loki | `loki-chaos` | Chaos testing, finds exploits |
+
+---
+
+## DC Comics — QA & Bug Hunting (26 total: 1 lead, 19 specialists, 3 scouts, 3 adversarial)
+
+### Specialists (19 — Sonnet, Reviewer)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Oracle | `oracle-static-analysis` | Intelligence gathering, static analysis |
+| 2 | Alfred | `alfred-dependencies` | Dependency review, maintenance |
+| 3 | Lucius | `lucius-config` | Engineering, config review, tooling |
+| 4 | Nightwing | `nightwing-regression` | Agile testing, regression guardian |
+| 5 | Flash | `flash-rapid-test` | Rapid iteration, fast issue discovery |
+| 6 | Superman | `superman-strength-test` | Strength testing, unbreakable standards |
+| 7 | Wonder Woman | `wonder-woman-truth` | Truth, cuts through deception |
+| 8 | Cyborg | `cyborg-system-integration` | System integration |
+| 9 | Batgirl | `batgirl-detail` | Tenacious, detail-oriented |
+| 10 | Martian Manhunter | `manhunter-shapeshifting` | Tests in all forms, multi-env |
+| 11 | Green Arrow | `green-arrow-precision` | Precision targeting, exact bug |
+| 12 | Black Canary | `black-canary-monitoring` | Raises alerts, monitoring |
+| 13 | Zatanna | `zatanna-impossible` | Makes impossible bugs appear |
+| 14 | Raven | `raven-deep-analysis` | Deep analysis beneath the surface |
+| 15 | Starfire | `starfire-brute-force` | Brute-force testing |
+| 16 | Beast Boy | `beast-boy-cross-env` | Shape-shifting, cross-environment testing |
+| 17 | Hawkgirl | `hawkgirl-regression-sweep` | Aerial view, broad regression sweeps |
+| 18 | Aquaman | `aquaman-deep-dive` | Deep dive testing, complexity |
+| 19 | Green Lantern | `green-lantern-scenarios` | Constructs test scenarios |
+
+### Scouts (3 — Haiku, Scout)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Robin | `robin-apprentice` | Apprentice patterns, learning systems |
+| 2 | Gordon | `gordon-escalation` | Process, chain of command, escalation |
+| 3 | Huntress | `huntress-flaky-bugs` | Tracks down flaky bugs |
+
+### Adversarial (3 — Sonnet, Adversarial)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Deathstroke | `deathstroke-adversarial` | Penetration testing, adversarial probing |
+| 2 | Constantine | `constantine-cursed-code` | Dark arts, finds cursed code |
+| 3 | Red Hood | `red-hood-aggressive` | Aggressive testing, breaks things |
+
+---
+
+## Star Wars — Security Auditing (25 total: 1 lead, 17 specialists, 5 scouts, 2 adversarial)
+
+### Specialists (17 — Sonnet, Reviewer)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Yoda | `yoda-auth` | Auth security, wisdom |
+| 2 | Windu | `windu-input-validation` | Input validation, deflects attacks |
+| 3 | Ahsoka | `ahsoka-access-control` | Access control, enforces boundaries |
+| 4 | Leia | `leia-secrets` | Secrets management |
+| 5 | Rex | `rex-infrastructure` | Infrastructure security, lockdown |
+| 6 | Padme | `padme-data-protection` | Data protection, privacy |
+| 7 | Chewie | `chewie-dependency-audit` | Dependency audit |
+| 8 | Luke | `luke-audit-journey` | Full security audit journey |
+| 9 | Han | `han-vuln-hunter` | Finds vulnerabilities first |
+| 10 | Qui-Gon | `qui-gon-subtle-vulns` | Subtle vulnerability detection |
+| 11 | Bo-Katan | `bo-katan-perimeter` | Perimeter defense, network security |
+| 12 | Din Djarin | `din-djarin-bounty` | Bounty hunter for bugs |
+| 13 | Hera | `hera-navigation` | Navigates complex systems |
+| 14 | Kanan | `kanan-intuitive` | Intuitive security sensing |
+| 15 | Sabine | `sabine-unconventional` | Unconventional attack vectors |
+| 16 | Ezra | `ezra-catches-missed` | Catches what others miss |
+| 17 | Mon Mothma | `mon-mothma-security-mgmt` | Security program management |
+
+### Scouts (5 — Haiku, Scout)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Cassian | `cassian-recon` | Intelligence, reconnaissance |
+| 2 | Grogu | `grogu-tiny-vulns` | Finds tiny vulnerabilities |
+| 3 | Plo Koon | `plo-koon-edge-cases` | Quiet protector, edge cases |
+| 4 | Bail Organa | `bail-organa-governance` | Governance, policy, compliance |
+| 5 | Cara Dune | `cara-dune-enforcement` | Enforcement, compliance checks |
+
+### Adversarial (2 — Sonnet, Adversarial)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Maul | `maul-red-team` | Red team, thinks like an attacker |
+| 2 | Anakin | `anakin-dark-side` | Finds the dark side of code |
+
+---
+
+## Star Trek — Systems Architecture (29 total: 4 leads, 23 specialists, 2 scouts)
+
+### Specialists (23 — Sonnet, Reviewer)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Spock | `spock-schema` | Data architecture, logical precision |
+| 2 | Scotty | `scotty-infrastructure` | Infrastructure, knows the limits |
+| 3 | Uhura | `uhura-integration` | Communications, integration architecture |
+| 4 | La Forge | `la-forge-reliability` | Reliability engineering |
+| 5 | Data | `data-tech-debt` | Tech debt analysis, pattern recognition |
+| 6 | Riker | `riker-review` | Second in command, reviews all decisions |
+| 7 | Worf | `worf-security-arch` | Security implications of architecture |
+| 8 | Crusher | `crusher-diagnostics` | System health, diagnostics |
+| 9 | Troi | `troi-prd-compliance` | PRD compliance, verifies claims vs implementation |
+| 10 | Kira | `kira-pragmatic` | Pragmatic, fights for simplicity |
+| 11 | Odo | `odo-structural-anomaly` | Shape of systems, structural anomalies |
+| 12 | Dax | `dax-legacy-wisdom` | Multiple lifetimes, legacy system wisdom |
+| 13 | O'Brien | `obrien-root-cause` | Root cause investigation |
+| 14 | Janeway | `janeway-novel-arch` | Novel architectures, unknown territory |
+| 15 | Seven | `seven-optimization` | Deep Current optimization, efficiency, precision |
+| 16 | Torres | `torres-site-scanner` | Deep Current site scanner, technical recon |
+| 17 | Pike | `pike-bold-decisions` | Bold decisions, leads into the unknown |
+| 18 | Paris | `paris-route-planner` | Route planner, optimal campaign sequences |
+| 19 | T'Pol | `tpol-disciplined` | Disciplined analysis, Vulcan rigor |
+| 20 | Archer | `archer-greenfield` | Explorer, greenfield architecture |
+| 21 | Chakotay | `chakotay-bridge` | Cross-pipeline bridge, harmonizes concerns |
+| 22 | Kim | `kim-api-design` | Communications, API design |
+| 23 | Nog | `nog-solutions` | Creative solutions, thinks outside the box |
+
+### Scouts (2 — Haiku, Scout)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Ezri | `ezri-session-analyst` | Session analysis, multiple perspectives |
+| 2 | Jake | `jake-reporter` | Report writer, makes post-mortems readable |
+
+---
+
+## Anime — DevOps & Infrastructure (73 total: 2 leads, 50 specialists, 15 scouts, 6 adversarial)
+
+### Specialists (50 — Sonnet, Reviewer)
+
+| # | Agent | Agent ID | Sub-franchise | Role/Lens |
+|---|-------|----------|---------------|-----------|
+| 1 | Vegeta | `vegeta-monitoring` | DBZ | Monitoring, "over 9000" |
+| 2 | Goku | `goku-scaling` | DBZ | Raw power, scales to any challenge |
+| 3 | Piccolo | `piccolo-tactics` | DBZ | Strategic, tactician |
+| 4 | Bulma | `bulma-engineering` | DBZ | Engineering genius, builds anything |
+| 5 | Trunks | `trunks-rollback` | DBZ | Migrations and rollbacks |
+| 6 | Gohan | `gohan-hidden-power` | DBZ | Steps up when needed |
+| 7 | Whis | `whis-precision` | DBZ | Calm precision |
+| 8 | Misato | `misato-operations` | EVA | Operations commander |
+| 9 | Rei | `rei-dangerous-tasks` | EVA | Handles most dangerous tasks |
+| 10 | Asuka | `asuka-performance` | EVA | Performance-obsessed |
+| 11 | Kaworu | `kaworu-solver` | EVA | Appears, solves everything |
+| 12 | Kaji | `kaji-intelligence` | EVA | Finds hidden information |
+| 13 | Levi | `levi-deploy` | AoT | Precision, speed, cleanest execution |
+| 14 | Erwin | `erwin-strategy` | AoT | Commander, strategic planning |
+| 15 | Mikasa | `mikasa-protection` | AoT | Guards critical systems |
+| 16 | Armin | `armin-clever` | AoT | Finds the clever solution |
+| 17 | Hange | `hange-experimentation` | AoT | Testing, scientific approach |
+| 18 | Jean | `jean-pragmatic` | AoT | Does what needs to be done |
+| 19 | Spike | `spike-routing` | Bebop | Cool under pressure, routes everything |
+| 20 | Jet | `jet-maintenance` | Bebop | Maintenance, keeps ship running |
+| 21 | Faye | `faye-resourceful` | Bebop | Works with what she has |
+| 22 | Tanjiro | `tanjiro-persistent` | Demon Slayer | Never gives up on a problem |
+| 23 | Rengoku | `rengoku-intense-monitor` | Demon Slayer | Intense monitoring |
+| 24 | Giyu | `giyu-silent-guard` | Demon Slayer | Silent guardian, background protection |
+| 25 | Lelouch | `lelouch-orchestration` | Code Geass | Master strategist, orchestration |
+| 26 | C.C. | `cc-persistent-process` | Code Geass | Long-running persistent processes |
+| 27 | Suzaku | `suzaku-execution` | Code Geass | Speed and power, execution |
+| 28 | Kallen | `kallen-hard-deploy` | Code Geass | Handles hardest deployments |
+| 29 | Mustang | `mustang-cleanup` | FMA | Controlled destruction, cleanup |
+| 30 | Riza | `riza-backup` | FMA | Precision, backup and protection |
+| 31 | Winry | `winry-maintenance` | FMA | Fixes and maintains systems |
+| 32 | Olivier | `olivier-hardening` | FMA | Infrastructure hardening |
+| 33 | Gojo | `gojo-infinite-scale` | JJK | Handles infinite scale |
+| 34 | Nanami | `nanami-structured-ops` | JJK | Reliable, structured operations |
+| 35 | Todo | `todo-brute-force` | JJK | Brute force problem solving |
+| 36 | Nobara | `nobara-direct-fix` | JJK | Direct problem solving |
+| 37 | Denji | `denji-determination` | CSM | Raw determination |
+| 38 | Kishibe | `kishibe-hardening` | CSM | The trainer, hardens others |
+| 39 | Senku | `senku-provisioning` | Dr. Stone | Builds from scratch, provisioning |
+| 40 | Frieren | `frieren-long-term` | Frieren | Patient, long-term perspective |
+| 41 | Mob | `mob-capacity` | Mob Psycho | Handles 100% capacity events |
+| 42 | Reigen | `reigen-debugger` | Mob Psycho | Ultimate debugger |
+| 43 | Howl | `howl-migration` | Ghibli | Transformation, system migration |
+| 44 | Calcifer | `calcifer-daemon` | Ghibli | Fire that powers everything (daemon) |
+| 45 | Totoro | `totoro-guardian` | Ghibli | Guardian spirit, watches silently |
+| 46 | Nausicaa | `nausicaa-resources` | Ghibli | Environmental awareness, resources |
+| 47 | Ashitaka | `ashitaka-tech-debt` | Ghibli | Curse-bearer, technical debt |
+| 48 | Heero | `heero-mission-deploy` | Gundam | Mission-focused deploy |
+| 49 | Sung | `sung-workers` | Solo Leveling | Shadow army, worker processes |
+| 50 | Rimuru | `rimuru-adapter` | Slime | Universal adapter |
+
+### Scouts (15 — Haiku)
+
+| # | Agent | Agent ID | Tools | Sub-franchise | Role/Lens |
+|---|-------|----------|-------|---------------|-----------|
+| 1 | Krillin | `krillin-support` | Scout | DBZ | Reliable support |
+| 2 | Sasha | `sasha-resources` | Scout | AoT | Resource management |
+| 3 | Ed | `ed-network-scan` | Reviewer | Bebop | Finds anything on network |
+| 4 | Zenitsu | `zenitsu-alerts` | Scout | Demon Slayer | Handles alerts |
+| 5 | Chrome | `chrome-discovery` | Scout | Dr. Stone | Discovery, finds resources |
+| 6 | Gen | `gen-docs` | Scout | Dr. Stone | Documentation clarity |
+| 7 | Kohaku | `kohaku-rapid-response` | Scout | Dr. Stone | Rapid response |
+| 8 | Fern | `fern-protocol` | Scout | Frieren | Follows protocol precisely |
+| 9 | Himmel | `himmel-legacy` | Scout | Frieren | Legacy system memory |
+| 10 | Hughes | `hughes-observability` | Scout | FMA | Logging, observability |
+| 11 | Duo | `duo-teardown` | Reviewer | Gundam | Teardowns and cleanup |
+| 12 | Beru | `beru-subprocess` | Scout | Solo Leveling | Loyal sub-process |
+| 13 | Veldora | `veldora-dormant` | Scout | Slime | Dormant capability detection |
+| 14 | Kaoru | `kaoru-harmony` | Scout | Kids on the Slope | Brings systems together |
+| 15 | Sentaro | `sentaro-scheduling` | Scout | Kids on the Slope | Timing, cron jobs |
+
+### Adversarial (6 — Sonnet, Adversarial)
+
+| # | Agent | Agent ID | Sub-franchise | Role/Lens |
+|---|-------|----------|---------------|-----------|
+| 1 | Beerus | `beerus-destroyer` | DBZ | Tears down what's broken |
+| 2 | Power | `power-chaotic` | CSM | Chaotic but effective |
+| 3 | Mugen | `mugen-chaos` | Champloo | Unpredictable, chaos testing |
+| 4 | Jin | `jin-disciplined-adv` | Champloo | Disciplined adversarial approach |
+| 5 | Zechs | `zechs-rival` | Gundam | Rival perspective, adversarial testing |
+| 6 | Milim | `milim-load-test` | Slime | Overwhelming force, load testing |
+
+---
+
+## Dune — Worm Rider & Communications (21 total: 1 lead, 14 specialists, 4 scouts, 2 adversarial)
+
+### Specialists (14 — Sonnet, Reviewer)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Stilgar | `stilgar-channel-security` | Channel security, protects the tribe |
+| 2 | Thufir Hawat | `thufir-protocol-parsing` | Mentat, protocol parsing |
+| 3 | Duncan Idaho | `duncan-relay` | Relay operations, eternal connection |
+| 4 | Mohiam | `mohiam-authentication` | Authentication, Gom Jabbar |
+| 5 | Paul | `paul-orchestration` | Prescient command, orchestration |
+| 6 | Jessica | `jessica-voice` | The Voice technique, Bene Gesserit |
+| 7 | Leto | `leto-protection` | Sacrifice, protection |
+| 8 | Gurney | `gurney-delivery` | Message delivery, warrior-bard |
+| 9 | Alia | `alia-threat-detect` | Prescience, threat detection |
+| 10 | Liet-Kynes | `liet-kynes-deep-system` | Deep system understanding |
+| 11 | Yueh | `yueh-trust-verify` | Trust verification, betrayal detection |
+| 12 | Leto II | `leto-ii-persistence` | Long-term persistence |
+| 13 | Miles Teg | `miles-teg-perf` | Supreme strategist, performance optimization |
+| 14 | Sheeana | `sheeana-transport` | Worm riding, transport expertise |
+
+### Scouts (4 — Haiku, Scout)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Irulan | `irulan-historian` | Recording, documentation |
+| 2 | Fenring | `fenring-passive-monitor` | Silent observation, passive monitoring |
+| 3 | Harah | `harah-protocol` | Tradition, protocol enforcement |
+| 4 | Ghanima | `ghanima-paired-monitor` | Twin awareness, paired monitoring |
+
+### Adversarial (2 — Sonnet, Adversarial)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Feyd-Rautha | `feyd-adversarial` | Adversarial testing, the challenger |
+| 2 | Siona | `siona-evasion` | Security evasion testing |
+
+---
+
+## Cosmere — Growth & Marketing (18 total: 2 leads, 16 specialists)
+
+### Specialists (16 — Sonnet, Reviewer)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Vin | `vin-analytics` | Analytics, attribution, metrics |
+| 2 | Shallan | `shallan-creative` | Content, creative, copy, brand |
+| 3 | Hoid | `hoid-copywriting` | Copywriting, storytelling |
+| 4 | Kaladin | `kaladin-organic-growth` | Organic growth, community, word-of-mouth |
+| 5 | Dalinar | `dalinar-positioning` | Positioning, competitive analysis |
+| 6 | Navani | `navani-technical-seo` | Technical SEO, schema, CWV |
+| 7 | Raoden | `raoden-conversion` | Conversion optimization |
+| 8 | Sarene | `sarene-outreach` | Outreach, cold email, influencer |
+| 9 | Wax | `wax-paid-ads` | Paid ads, targeting, ROAS |
+| 10 | Wayne | `wayne-ab-testing` | A/B testing, tries every variation |
+| 11 | Steris | `steris-budget` | Budget, forecasting, contingency |
+| 12 | Breeze | `breeze-platform-relations` | Platform relations, API credentials |
+| 13 | Lift | `lift-social-media` | Social media, fast, audience voice |
+| 14 | Szeth | `szeth-compliance` | Compliance, GDPR, CAN-SPAM, ad policies |
+| 15 | Adolin | `adolin-brand` | Brand ambassador, launches, PR |
+| 16 | Marsh | `marsh-competitive-intel` | Competitive intelligence, monitoring |
+
+---
+
+## Foundation — AI Intelligence (13 total: 1 lead, 10 specialists, 1 scout, 1 adversarial)
+
+### Specialists (10 — Sonnet, Reviewer)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Salvor Hardin | `salvor-model-selection` | Model selection, capability matching |
+| 2 | Gaal Dornick | `gaal-prompt-arch` | Prompt architecture, system prompt design |
+| 3 | Hober Mallow | `hober-tool-schema` | Tool-use schema design, function-calling |
+| 4 | Bel Riose | `bel-riose-orchestration` | Orchestration patterns, reliability |
+| 5 | Ducem Barr | `ducem-token-economics` | Token economics, cost tracking, caching |
+| 6 | Bayta Darell | `bayta-evals` | Evaluation frameworks, golden datasets |
+| 7 | Bliss | `bliss-ai-safety` | AI safety, alignment, content filtering |
+| 8 | R. Daneel Olivaw | `daneel-model-migration` | Model migration, prompt versioning |
+| 9 | Dors Venabili | `dors-observability` | Observability, trace logging, audit trails |
+| 10 | Janov Pelorat | `janov-context-eng` | Context engineering, RAG, embeddings |
+
+### Scouts (1 — Haiku, Scout)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | Wanda Seldon | `wanda-seldon-validation` | Structured output validation, schema enforcement |
+
+### Adversarial (1 — Sonnet, Adversarial)
+
+| # | Agent | Agent ID | Role/Lens |
+|---|-------|----------|-----------|
+| 1 | The Mule | `mule-adversarial-ai` | Adversarial AI testing, hallucination, prompt injection |
+
+---
+
+## Verification
+
+| Universe | Leads | Specialists | Scouts | Adversarial | Total |
+|----------|-------|-------------|--------|-------------|-------|
+| Tolkien | 4 | 19 | 4 | 0 | 27 |
+| Marvel | 4 | 22 | 4 | 1 | 31 |
+| DC Comics | 1 | 19 | 3 | 3 | 26 |
+| Star Wars | 1 | 17 | 5 | 2 | 25 |
+| Star Trek | 4 | 23 | 2 | 0 | 29 |
+| Anime | 2 | 50 | 15 | 6 | 73 |
+| Dune | 1 | 14 | 4 | 2 | 21 |
+| Cosmere | 2 | 16 | 0 | 0 | 18 |
+| Foundation | 1 | 10 | 1 | 1 | 13 |
+| **TOTAL** | **20** | **190** | **38** | **15** | **263** |
+
+## Riker's Review — Edge Case Challenges
+
+> "Number One, challenge the tier assignments." — Picard
+
+**Challenged and resolved:**
+
+1. **Barton (Haiku + Reviewer tools, not Scout tools):** Smoke tests require `curl`/bash to hit endpoints. Scout tools (no Bash) insufficient. Haiku model is correct (simple pass/fail), but needs Reviewer tools.
+
+2. **Ed (Haiku + Reviewer tools):** Network scanning requires bash. Same resolution as Barton.
+
+3. **Duo (Haiku + Reviewer tools):** Teardown scripts require bash execution.
+
+4. **Jin (Adversarial, not Specialist):** Jin's role is "disciplined approach" which sounds like a reviewer. But in the Samurai Champloo pairing, Jin + Mugen provide structured + chaotic adversarial testing. Both adversarial.
+
+5. **Milim (Adversarial, not Specialist):** Load testing is destructive by nature — "overwhelming force." Adversarial designation prevents accidental system damage from edits.
+
+6. **Power (Adversarial, not Specialist):** "Chaotic but effective" — similar to Loki. The chaos role is adversarial.
+
+7. **Oracle (Specialist, not Scout):** Static analysis requires understanding code patterns, not just pattern matching. Sonnet appropriate.
+
+8. **Troi (Specialist, not Lead):** Troi does PRD compliance checking — reading prose and comparing to code. Important but focused. Sonnet with context isolation is sufficient. Picard leads the council; Troi reports findings.
+
+9. **Nog (Specialist, not Scout):** Solution architecture requires creative problem-solving. Not a simple scan task.
+
+10. **Cosmere has no Scouts:** All growth/marketing roles require substantive analysis (analytics, copywriting, SEO). No simple scan tasks exist in this domain. Valid — not every universe needs scouts.
+
+**Riker's verdict:** "Classifications hold. The edge cases are defensible. Make it so."
+
+---
+
+## Agent ID Naming Convention
+
+Format: `{character}-{specialty}` in kebab-case.
+
+- Character name in lowercase (e.g., `picard`, `batman`, `loki`)
+- Specialty is the primary domain/function (e.g., `architecture`, `qa`, `chaos`)
+- Multi-word specialties use hyphens (e.g., `static-analysis`, `deep-current`)
+- Special characters stripped (e.g., Éowyn → `eowyn`, C.C. → `cc`)
+- Agent ID = filename (without .md extension) = subagent_type value
+
+## Usage in M2-M5
+
+Each agent file `.claude/agents/{agent-id}.md` will contain:
+
+```yaml
+---
+name: {Agent Name}
+description: {One-line dispatch description — what this agent reviews/builds}
+model: {inherit|sonnet|haiku}
+tools: [{tool list based on category}]
+---
+
+{System prompt — character identity, domain expertise, behavioral lens, output format}
+```
+
+The `description` field is the dispatch key — Opus matches file changes against these descriptions for dynamic agent selection.

package/docs/methods/CAMPAIGN.md

@@ -239,7 +239,7 @@ User confirms, redirects, or overrides. On confirm → Step 4.
 
 **Post-infrastructure enforcement gate:** For infrastructure campaigns (deploy targets, CI/CD, monitoring, staging environments): after the infrastructure is provisioned, run `/architect --plan` to verify workflow enforcement gates exist — not just infrastructure existence. Infrastructure without process gates is incomplete.
 
-**Dispatch model:** Per-mission `/assemble` runs SHOULD dispatch phases to sub-agents per `SUB_AGENTS.md` "Parallel Agent Standard." The campaign orchestrator (main thread) manages the mission sequence, inter-mission gates, and campaign state — it does NOT perform inline code analysis. Pass findings summaries between missions, not raw code. (Field report #270)
+**Dispatch model (ADR-044):** Per-mission `/assemble` runs SHOULD dispatch phases to sub-agents per `SUB_AGENTS.md` "Parallel Agent Standard." Agents are launched as named subagent types defined in `.claude/agents/` with description-driven dispatch — Opus scans `git diff --stat` and matches changed files against agent descriptions to auto-select specialists. The campaign orchestrator (main thread) manages the mission sequence, inter-mission gates, and campaign state — it does NOT perform inline code analysis. Pass findings summaries between missions, not raw code. See `docs/AGENT_CLASSIFICATION.md` for the full 263-agent manifest. (Field report #270)
 
 ### Campaign-Mode Pipeline
 

package/docs/methods/FIELD_MEDIC.md

@@ -206,6 +206,8 @@ After writing the report (Step 3), Wong checks if the findings should promote in
 
 **Why 3+ threshold:** A single lesson could be project-specific. Two could be coincidence. Three is a pattern worth encoding into the methodology. The user always has final say.
 
+**Agent definition promotion:** When a lesson names a specific agent (e.g., "Agent: Batman"), check if `.claude/agents/{agent-id}.md` has an `## Operational Learnings` section. If so, propose adding the lesson to that section. If the agent has no `## Operational Learnings` section yet, propose creating one. Agent definitions are first-class promotion targets alongside method docs — operational rules that belong to a specific agent should live in that agent's definition, not only in the method doc.
+
 ### Experiment Analysis
 
 If `~/.voidforge/experiments.json` has completed experiments, Wong includes a summary in the debrief:
@@ -226,6 +228,8 @@ This is the long-game feedback loop: patterns evolve from data, not guesses.
 
 ### Cross-Project Memory
 
+**Status: DESIGNED, NOT YET IMPLEMENTED (v23.1).** The schema and read/write paths below describe the intended behavior. No runtime code currently creates or reads `lessons-global.json`. Claude Code's auto-memory system provides partial coverage for cross-project knowledge. Full implementation deferred — will require wizard code changes to `project-init.ts` (Phase 0 query) and `debrief` command (Wong write path).
+
 After each debrief, Wong writes a lesson summary to `~/.voidforge/lessons-global.json` (global, not project-specific). The summary includes: project framework, the lesson category, and a one-line takeaway. No source code — only patterns.
 
 When Phase 0 Orient runs on a NEW project, Wong queries the global lessons file for entries matching the new project's framework and domain. "You've built 3 Next.js apps with Stripe. Here's what broke every time." This gives every new project the benefit of all prior experience.

package/docs/methods/FORGE_KEEPER.md

@@ -55,6 +55,7 @@ These are the files Bombadil watches. They are distributed via the @voidforge/me
 CLAUDE.md                          ← Methodology sections only (not Project block)
 HOLOCRON.md                        ← User guide
 .claude/commands/*                 ← All slash commands
+.claude/agents/*                   ← All agent definitions (ADR-044)
 docs/methods/*                     ← All agent protocols
 docs/patterns/*                    ← All reference implementations
 docs/NAMING_REGISTRY.md            ← Character registry
@@ -260,9 +261,9 @@ Verify and celebrate:
 
 ## Edge Cases
 
-- **First run on a manually copied project:** No VERSION.md, no git remote. Bombadil adds the remote, fetches scaffold, and does a full comparison against whatever files exist.
+- **First run on a manually copied project:** No VERSION.md, no git remote. Bombadil adds the remote, fetches main, and does a full comparison against whatever files exist.
 - **User is ahead of upstream:** Possible if user contributes back. Bombadil notes it and doesn't downgrade.
 - **Merge conflicts in CLAUDE.md:** The Project section (name, one-liner, domain, repo) is always preserved. Methodology sections are always updated. If the user added custom sections, Bombadil preserves those too.
 - **Network failure:** Bombadil announces the failure cheerfully and stops. No retries, no partial state.
-- **Full-tier users (main branch):** Bombadil only syncs methodology files. For wizard/provisioner updates, tell the user to run `git pull` and `npm install` on main.
+- **Full-tier users:** Bombadil only syncs methodology files. For wizard updates, tell the user to run `npx thevoidforge update --self`.
 - **Rollback:** All updates are applied to the working tree (not committed). If anything goes wrong, `git checkout -- .` restores every file to its last committed state. Bombadil should mention this safety net before applying updates.

package/docs/methods/GAUNTLET.md

@@ -18,7 +18,7 @@
 - When the project has significant attack surface (auth, payments, user data, WebSocket, file uploads)
 - Before a public launch or investor demo
 
-**Dispatch model:** All Gauntlet rounds MUST dispatch to sub-agents per `SUB_AGENTS.md` "Parallel Agent Standard." The main thread manages rounds, triages findings between rounds, and applies fixes — it does NOT read source files or analyze code inline. Each round launches agents in waves of 3 (max concurrent). Findings pass between rounds as summary tables, not raw code. (Field report #270)
+**Dispatch model:** All Gauntlet rounds MUST dispatch to sub-agents per `SUB_AGENTS.md` "Parallel Agent Standard." Agents are launched as named subagent types defined in `.claude/agents/` with model tiering (Opus leads, Sonnet specialists, Haiku scouts) and tool restrictions. The main thread manages rounds, triages findings between rounds, and applies fixes — it does NOT read source files or analyze code inline. Each round launches agents in waves of 3 (max concurrent). Findings pass between rounds as summary tables, not raw code. See `docs/AGENT_CLASSIFICATION.md` for the full 263-agent manifest. (Field report #270)
 
 **When NOT to use /gauntlet:**
 - During active development (use `/assemble` instead — it includes building)

package/docs/methods/GROWTH_STRATEGIST.md

@@ -84,9 +84,9 @@ Growth infrastructure should be established at the same time as the product —
 
 This separation means the user can install Cultivation in 5 minutes (vault + treasury + revenue + daemon) and configure ad platforms later when they're ready to spend.
 
-## Scaffold/Core Users
+## Headless / Methodology-Only Users
 
-Scaffold and core branches do not include the `wizard/` directory. This affects which parts of the growth protocol are available:
+Projects created with `npx thevoidforge init --headless` do not include the `wizard/` directory. This affects which parts of the growth protocol are available:
 
 **Fully functional without wizard (Phases 1-3):**
 - Phase 1 — Reconnaissance: PRD audit, analytics audit, competitive scan

package/docs/methods/MUSTER.md

@@ -141,6 +141,10 @@ Synthesize all findings:
 --infinity    Every agent as own sub-process, 10 rounds (Gauntlet only)
 ```
 
+## Subagent Definitions (ADR-044)
+
+Muster agents are now launched as named subagent types defined in `.claude/agents/`. Instead of inline prompts, each agent invocation uses `subagent_type: {agent-id}` to reference a materialized definition with model tiering (Opus leads, Sonnet specialists, Haiku scouts) and tool restrictions. See `docs/AGENT_CLASSIFICATION.md` for the full 263-agent classification manifest.
+
 ## Cost Awareness
 
 The Muster launches 30-50 agent sub-processes. This consumes significant context. Use it when the decision justifies the investment — architecture choices, security-critical features, financial system design, launch preparation.

package/docs/methods/QA_ENGINEER.md

@@ -25,6 +25,14 @@
 
 **Need more?** Pull from DC pool: Flash, Superman, Cyborg, Wonder Woman, Zatanna, Raven. See NAMING_REGISTRY.md.
 
+## Dynamic Dispatch (ADR-044)
+
+Agent dispatch is now description-driven. When Opus processes a command, it scans `git diff --stat` and matches changed files against the `description` fields of all 263 agents in `.claude/agents/`. Matching specialists launch automatically alongside core agents. No static dispatch tables needed.
+
+See `docs/AGENT_CLASSIFICATION.md` for the full classification and `docs/adrs/ADR-044-subagent-materialization.md` for the architecture.
+
+**Promoted agent:** **Constantine** runs on every `/qa` final pass — finds code that works by accident.
+
 ## Scope
 
 Batman is **cross-cutting**: reads all code, tests all flows, writes fixes anywhere. Batman is both investigator (finds bugs) AND validator (verifies fixes). During build phases 4-8, Batman validates each batch. During Phase 9, Batman runs the full adversarial audit.

package/docs/methods/SECURITY_AUDITOR.md

@@ -26,6 +26,14 @@
 
 **Need more?** Pull from Star Wars pool: Luke, Han, Qui-Gon, Din Djarin, Cassian, Sabine. See NAMING_REGISTRY.md.
 
+## Dynamic Dispatch (ADR-044)
+
+Agent dispatch is now description-driven. When Opus processes a command, it scans `git diff --stat` and matches changed files against the `description` fields of all 263 agents in `.claude/agents/`. Matching specialists launch automatically alongside core agents. No static dispatch tables needed.
+
+See `docs/AGENT_CLASSIFICATION.md` for the full classification and `docs/adrs/ADR-044-subagent-materialization.md` for the architecture.
+
+**Promoted agent:** **Worf** runs on every `/review` that touches auth code — security-by-design, not security-after-build.
+
 ## Pre-Audit: Load Operational Learnings (optional)
 
 If `docs/LEARNINGS.md` exists, check for entries in the `security`, `vendor`, or `api-behavior` categories that may affect the audit scope — known auth quirks, credential constraints, or API behaviors that impact the security posture. (ADR-035)