adaptive-team-consult

name	adaptive-team-consult
description	Consult the team on a topic — PM selects the right specialists, they brief and discuss. No code is written.
argument-hint	<topic or question>

Adaptive Team Consult

PM brings in the specialists whose lenses fit the topic and facilitates a plain-English discussion. No code is written; the output is shared understanding and a direction.

Pipeline

Step 0: PM Recovery / Grounding

PM reads, in order:

1. .claude/rules/adaptive-team-rules.md
2. .claude/adaptive-team-state/current-session.md (create from template if missing)
3. .claude/adaptive-team-learned/pm-lessons.md
4. TaskList

Step 1: Classify the Topic

PM identifies topic signals and selects reviewers per the Consult Mode rubric in the rules:

Signal	Spawn
Schema, query, migration, index, graph	database
Prompt, tool use, model choice, agent design, eval	llm-expert
Auth, permissions, authz, secrets	architect
Test strategy, coverage, flake	sdet
Architecture / coherence	architect
Ambiguous	architect as default

Default minimum: architect. Spawn only those the signal warrants — do not invite everyone by habit.

Step 2: Create or Join Team

If a team exists, join it — do not recreate. Otherwise TeamCreate with a descriptive name. Update current-session.md.

Step 3: Spawn Selected Reviewers

Use the deterministic names: architect, sdet, llm-expert, database. Each completes its startup sequence.

Step 4: Facilitate the Consult

PM forwards the user's topic to spawned reviewers and asks each for:

• Their plain-English read
• Risks or constraints specific to their lens
• Recommended direction (not code)

Each reviewer writes a full analysis to .claude/adaptive-team-reviews/consult/<short-slug>/<reviewer>.md and returns a short summary to PM. PM synthesizes for the user by showing one paragraph per reviewer, verbatim from the summary line — PM does not re-author.

Step 5: Decide or Iterate

User decides the direction, asks follow-ups, or requests /adaptive-team-implement to build.

Step 6: Update Session State

PM updates current-session.md with any key decisions and the set of spawned reviewers (which remain persistent for this session).

When to Use

• Approach or design questions before implementation
• Choosing between options (model, schema, framework)
• Understanding risk and tradeoffs on a specific topic

When NOT to Use

• Actual implementation → /adaptive-team-implement
• Capturing a lesson from something that went wrong → /adaptive-team-learning-moment