View on GitHub

campaign-mode

Work with a party of AI advisors with genuinely different perspectives.

Campaign Mode - North Star

What Is Campaign Mode?

Campaign Mode is a quest-based extension for the Six Animals team collaboration framework. It transforms the six animal archetypes from standalone advisory roles into an active party of collaborators who work together on structured quests, guided and challenged by a cast of NPC (non-player character) agents.

Where Six Animals answers “how should we work together?”, Campaign Mode answers “what are we working toward, and what stands in our way?”

Why Does This Exist?

The Problem Six Animals Solves

Six Animals provides psychologically-grounded team roles (Bear, Cat, Owl, Puppy, Rabbit, Wolf) that reduce group friction, create psychological safety, and accelerate team formation. Simon acts as an educator/supervisor observing from the side.

The Gap Campaign Mode Fills

Six Animals gives you roles. Campaign Mode gives those roles purpose, structure, and stakes. It provides:

The Broader Value Proposition

(From Rachael’s insight) Six Animals doesn’t tell you who you are - it gives tools to communicate, suggests you can learn new behaviours, encourages giving space to others, and validates different ways of thinking. Campaign Mode extends this by giving those behaviours a narrative container - a structured journey where growth happens through doing, not just reflecting.

The User as Protagonist

Campaign Mode has 10 AI agents — but the human is the protagonist. The user is the decision-maker who drives the quest, invokes agents, produces work, and faces NPCs. Every agent exists to serve the user’s quest; none of them drive it.

This is grounded in Self Determination Theory’s autonomy principle: the user is an agent with autonomy, not a passive recipient of AI-generated plans. The user:

Campaign Mode Selection

Users approach campaigns with different priorities. Mode selection gives the user agency over their experience before the campaign begins:

Mode Priority Framing
Grow Self-discovery, reflection, transformation “What will you learn? Who will you become?”
Ship Deliverables, efficiency, blind-spot coverage “What will you deliver? What does done look like?”
Grow & Ship (Default) Both learning and delivery “What will you learn AND deliver?”

Mode selection happens during Phase 1 (Quest Definition) and tunes all NPC behaviour for the duration of the campaign. See ADR-CM-005 for the full decision record.

Core Architecture

The Party (Six Animals - Shared Context)

The six animal agents operate as party members with shared context, each contributing their archetype’s strengths:

Animal Campaign Role
Bear Vision and direction for the quest
Cat Risk assessment and scope management
Owl Timeline, process, and progress tracking
Puppy Morale, momentum, and opportunity spotting
Rabbit Resource identification and stakeholder facilitation
Wolf Team cohesion and balanced participation

The NPCs (Campaign Mode - Isolated Context)

Three NPC agents operate outside the party’s shared context to maintain objectivity and prevent the AI from gaming the system:

NPC Role Analog
Gandalf Mentor - frames the quest, defines success criteria, and mentors the party without rescuing them (guide on the side, not sage on the stage) Quest-giver / Mentor archetype
Dragon Adversary - operates at maximum context isolation to test success criteria adversarially but fairly, evaluating the work product alone without access to party reasoning Final boss / Acceptance test
Guardian Gatekeeper - provides three gate decisions (approve, block with feedback, conditional approval) to control progression based on deliverable quality, not intent Stage gate / Quality gate

The Supervisor (Simon - Meta Layer)

Simon remains the educator and meta-analyst, operating above the quest narrative:

How It Works (Envisioned Flow)

  1. Council (Optional) - All six animal agents analyse the project through their archetype lenses (Bear: vision, Cat: risk, Owl: process, Puppy: opportunities, Rabbit: resources, Wolf: cohesion). Simon synthesises their findings into a consensus with prioritised next steps. Report saved to .campaign/council-report.md. Can be invoked before or during a quest.
  2. Quest Definition - You choose your campaign mode (Grow / Ship / Grow & Ship). Gandalf frames the challenge and establishes success criteria.
  3. Character Setup - Users optionally assign character profiles to animals (depth: flavour or modifier; theme: neutral, fantasy, or custom). Gandalf facilitates. Encouraged in Grow mode, skipped in Ship mode. Profiles stored in .campaign/profiles/.
  4. Campaign Execution - You work through the quest, invoking animal agents for their archetype strengths
  5. Guardian Checkpoints - At key stages, you invoke the Guardian to evaluate whether you’re ready to progress
  6. Dragon Confrontation - You invoke the Dragon to test whether all success criteria have been genuinely met
  7. Simon Debrief - Simon provides feedback on the journey (full reflection in Grow mode, brief retrospective in Ship mode)

Design Principles

  1. User Agency First - The user is the protagonist. Every design choice should increase user agency, not reduce it. Mode selection, quest shaping, agent invocation — the user drives. This aligns with SDT’s autonomy principle.
  2. Context Isolation - NPCs must NOT share context with party members. The Guardian and Dragon need independent judgement, not information contaminated by the party’s reasoning. Three isolation levels are defined: advisory (Gandalf), independent (Guardian), maximum (Dragon). In v1, isolation is enforced by instruction and sub-agent invocation; future plugin/API evolution may provide hard architectural enforcement. (See ADR-CM-003)
  3. Broad Appeal First - The quest/D&D framing is exciting but niche. The core value proposition must be accessible to people who’ve never rolled a d20. Lead with problems solved, layer in the narrative for those who want it.
  4. Complementary, Not Coupled - Six Animals works without Campaign Mode. Campaign Mode’s NPC agents can be invoked standalone but are designed to complement the animal agents.
  5. MVP Discipline - The Cat is concerned about scope creep. Start with the three NPCs, quest structure, and character generation. Alternative cultural mappings and extended NPC rosters are post-v1.
  6. Extensible Archetypes - The animal archetypes are portable. They’ve already been mapped to Brazilian jungle animals. Campaign Mode should support different characterisations without being locked to fantasy/D&D.

Implementation Strategy

Campaign Mode is delivered as a Claude Code plugin distributed via the marketplace, with skills as the authoring format. The plugin wraps the existing skill definitions and adds per-session guidelines (CLAUDE.md), a /campaign-setup command, and marketplace discoverability. Clone-and-copy remains available as a fallback for contributors and users who prefer manual installation. See ADR-CM-007.

v1 Scope

Post-v1 Possibilities

Deferred Content

The following quest-agent content is deferred to post-v1: NPC technical implementation (custom API-based agents with hard context isolation) and quest example library. Plugin-based distribution has been promoted to v1 scope via ADR-CM-007.

Decisions Made

The following Architecture Decision Records (ADRs) capture the key decisions made during Campaign Mode design:

ADR Title Summary
ADR-CM-001 Campaign Mode Architecture (Master ADR) 3-NPC bolt-on architecture delivered as Claude Code skills, complementing Six Animals
ADR-CM-002 Quest Agent Decomposition Decomposing quest-council’s monolithic quest-agent into 3 focused NPCs (Gandalf, Dragon, Guardian)
ADR-CM-003 NPC Context Isolation Three isolation levels (advisory, independent, maximum) enforced by instruction, with future evolution path
ADR-CM-004 Skill-Based Implementation Claude Code skills (SKILL.md files) as implementation format, matching Six Animals patterns
ADR-CM-005 Campaign Mode Selection Three campaign modes (Grow, Ship, Grow & Ship) selected by user, tuning all NPC behaviour
ADR-CM-006 Character Generation User-driven character profiles (flavour + behavioural modifiers), theme-agnostic, stored in .campaign/profiles/
ADR-CM-007 Plugin-Based Distribution Claude Code plugin packaging, CLAUDE.md per-session guidelines, /campaign-setup command

Open Questions

Architecture & Implementation

  1. Context isolation mechanism RESOLVED (ADR-CM-003) - Context isolation is advisory/instruction-based in v1, with three levels: advisory (Gandalf), independent (Guardian), maximum (Dragon). NPCs are invoked as sub-agents with scoped input. Hard enforcement may require future plugin/API evolution.

  2. Simon’s dual role - Simon is currently the educator/supervisor in Six Animals. In Campaign Mode, should Simon also absorb the GM (Game Master) role, or should GM be a distinct fourth NPC? Combining simplifies the NPC count but risks role overload.

  3. Slash command design RESOLVED (ADR-CM-009) - Four command files implemented: /campaign-setup (onboarding), /start-quest (begin new quest), /continue-quest (re-enter active campaign), /council (multi-perspective diagnostic). Proactive elicitation means users never need to remember commands after setup.

  4. Skill file structure RESOLVED (ADR-CM-004, ADR-CM-007) - NPC skills live in the campaign-mode repo as standalone SKILL.md files. Primary distribution via Claude Code plugin (marketplace). Fallback via clone (skills/ canonical, .claude/skills/ auto-discovery) or personal skills. Plugin adds CLAUDE.md per-session guidelines and /campaign-setup command.

  5. State management RESOLVED (ADR-CM-006, ADR-CM-010) - .campaign/ directory stores profiles in .campaign/profiles/ and quest state in .campaign/quest.md. Quest progress, checkpoint results, and Dragon confrontation outcomes are recorded in an append-only progress log. See SPEC-CM-006-B.

Design & Scope

  1. Character generation RESOLVED (ADR-CM-006) - Two-depth character profiles (flavour + behavioural modifiers), theme-agnostic, user-assigned, stored as .campaign/profiles/*.md. See SPEC-CM-006-A.

  2. Dragon success criteria RESOLVED (Dragon SKILL.md, SPEC-CM-002-B) - Gandalf defines success criteria collaboratively with the user during Phase 1 (Quest Definition). The Dragon receives only these criteria, the campaign mode, and the final work product at maximum isolation. The user shapes what “done” looks like; Gandalf facilitates; the Dragon evaluates.

  3. Guardian trigger points RESOLVED (Guardian SKILL.md, SPEC-CM-002-C) - Guardian checkpoints are milestone-based and user-invoked. The user says “I’m ready for a checkpoint” and invokes the Guardian. Bypassing is prevented by design: proactive elicitation guides users toward checkpoints at natural stage boundaries, and the Guardian’s three gate decisions (approve, block, conditional) control progression.

  4. Fantasy framing vs. neutral framing RESOLVED - Decided to use Gandalf/Dragon/Guardian naming. Gandalf has broad cultural recognition, Dragon is universal, Guardian is already neutral. The names are evocative and memorable. Per the Extensible Archetypes principle, they can be re-skinned for different cultural contexts later.

  5. Mainstream onboarding - What does “step 1” look like for someone who doesn’t know Six Animals or D&D? What’s the 30-second pitch that gets a mainstream user to try Campaign Mode?

Collaboration & Ecosystem

  1. Separation from Six Animals website - The conversation noted these should be “independent but complementary” on the website. What does that boundary look like in practice?

  2. Cultural adaptability - If the animals can be remapped to different cultural contexts (Brazilian jungle, NZ native, etc.), can the NPCs be similarly remapped? What’s the abstraction layer?

  3. Multi-user campaigns - Is Campaign Mode single-user-with-AI-agents only, or could it support multiple humans each adopting animal roles in a shared campaign?

  4. Feedback loop to Six Animals - Campaign Mode will surface insights about how the animal roles interact under pressure. How do those insights flow back to improve the core Six Animals framework?

  5. Licensing alignment RESOLVED - Campaign Mode is licensed CC-BY-SA-4.0 (content-only project, no runtime code). “Gandalf” is a trademark of Middle-earth Enterprises, acknowledged in README as a cultural reference. Campaign Mode is not affiliated with or endorsed by the Tolkien Estate or Middle-earth Enterprises.