View on GitHub

campaign-mode

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

ADR-CM-011: Council Feature

Field Value
Decision ID ADR-CM-011
Type Child ADR
Parent ADR ADR-CM-001
Initiative Campaign Mode
Proposed By Chris Barlow
Date 2026-02-14
Status Accepted

WH(Y) Decision Statement

In the context of the gap between campaign setup and quest commitment — after /campaign-setup creates infrastructure, the user’s only option is to immediately start a quest, with no way to first understand the project from multiple perspectives,

facing users needing to understand their project’s state, strengths, and gaps before choosing a quest direction — particularly in unfamiliar codebases or when the right next step is unclear,

we decided for a /council command that sequentially invokes all six animal agents to independently analyse the project through their archetype lenses, followed by Simon synthesising findings into a consensus with prioritised next steps, persisted as .campaign/council-report.md,

and neglected parallel Task sub-agents (analysis invisible to user, loses the conversational feel), Gandalf-only analysis (loses diverse perspectives from six archetypes), and mandatory pre-quest Council (violates user agency — the Council must be optional),

to achieve a visible multi-perspective diagnostic where each animal speaks in character, a persistent synthesis artefact that can inform quest definition, and seamless integration with campaign setup as an on-ramp to quest commitment,

accepting that sequential animal invocation produces long output, the feature requires Six Animals to be installed, and the council report is overwritten on re-invocation (no history).


Context

Campaign Mode has a gap between setup and quest commitment. After /campaign-setup creates infrastructure (CLAUDE.md, .campaign/ directory), the user’s only transition is to start a quest — which immediately commits to a direction via Gandalf’s quest framing. There is no way to first survey the project from multiple perspectives before choosing what to work on.

The Council fills this gap: all six animal agents independently analyse the project repository through their archetype lenses (Bear: vision, Cat: risk, Owl: process, Puppy: opportunities, Rabbit: resources, Wolf: cohesion), then Simon synthesises their findings into a consensus with recommended next steps. The user can then optionally transition into quest definition with Gandalf, or take the analysis standalone.

This also addresses a bug in campaign-setup.md: Step 4 ends with a slash command reference (Run /start-quest) rather than proactive elicitation via AskUserQuestion, violating the pattern established in ADR-CM-008.

Options Considered

Option 1: Sequential In-Character Analysis (Selected)

Each animal speaks in the main conversation with speaker tags, followed by Simon’s synthesis. User sees the full diagnostic unfold.

Pros: Visible, engaging, conversational; user sees each perspective as it’s delivered; consistent with agent identity conventions Cons: Long sequential output; user must wait for all six analyses

Option 2: Parallel Sub-Agent Analysis (Rejected)

Launch six Task sub-agents in parallel, collect results, present a summary.

Why rejected: The analysis is invisible to the user. They see only the final summary, losing the engaging multi-perspective conversation. The Council’s value is partly in watching diverse viewpoints emerge.

Option 3: Gandalf-Only Analysis (Rejected)

Gandalf surveys the project and frames potential quest directions.

Why rejected: Gandalf is a mentor, not a diagnostic tool. The Council’s value comes from six distinct archetype lenses — Bear sees vision gaps that Cat sees as risks, Puppy sees opportunities that Owl structures. A single perspective, however wise, misses this diversity.

Option 4: Mandatory Pre-Quest Council (Rejected)

Require a Council session before any quest can begin.

Why rejected: Violates the User as Protagonist principle. The Council must be optional — some users know exactly what they want to work on and shouldn’t be forced through a diagnostic step.


Dependencies

Relationship ADR ID Title Notes
Extends ADR-CM-008 Proactive Elicitation Council uses AskUserQuestion for transitions; fixes campaign-setup elicitation gap
Extends ADR-CM-009 Quest Entry Commands Council provides an additional on-ramp between setup and quest commitment
Relates To ADR-CM-001 Campaign Mode Architecture Adds optional pre-quest diagnostic to campaign lifecycle
Relates To ADR-CM-010 Quest State Tracking Council report stored in .campaign/ alongside quest state

References

Reference ID Title Type Location
REF-001 Campaign Lifecycle Spec Internal Spec SPEC-CM-001-B
REF-002 Campaign State Directory Spec Internal Spec SPEC-CM-006-B
REF-003 Proactive Elicitation ADR Internal ADR ADR-CM-008
REF-004 Quest Entry Commands ADR Internal ADR ADR-CM-009

Governance

Review Board Date Outcome Action Review Cadence Next Review
Quarterly

Status History

Status Approver Date
Proposed Chris Barlow 2026-02-14
Accepted Chris Barlow 2026-02-14