Decision Audit Trail¶
Skills used: CONCLAVE + MIRROR + ECHO
Debate it. Stress-test it. Record it. The bulletproof decision pipeline.
The Problem¶
Big architectural decisions deserve more than a gut feeling. But most teams either:
- Decide too fast — first idea wins, nobody challenges it
- Decide too slow — endless meetings, analysis paralysis, decision by committee
You need a process that's thorough but fast. Rigorous but practical.
The Pipeline¶
Step 1: CONCLAVE — Let Them Fight¶
9 voices. 2 rounds. The minimalist argues for simplicity. The systems thinker maps the dependencies. The chaos agent breaks your assumptions. The builder asks if you can actually maintain it.
You get a synthesized recommendation with dissenting opinions preserved.
Time: 2 minutes.
Step 2: MIRROR — Stress-Test the Winner¶
Take CONCLAVE's recommendation and throw it in the mirror:
MIRROR picks 3 challenge frameworks and genuinely argues against the recommendation:
- Hidden Costs: Monorepo tooling overhead, CI complexity, blame diffusion
- Failure Modes: One broken build blocks everyone, merge conflicts at scale
- Reversibility: Splitting a monorepo later is painful — how painful?
If the recommendation survives MIRROR, it's solid. If it doesn't, you just saved yourself months of regret.
Time: 1 minute.
Step 3: ECHO — Carve It in Stone¶
The decision survived. Record it:
Claude captures:
- Decision: Monorepo with Turborepo
- Reasoning: CONCLAVE synthesis + MIRROR survival
- Alternatives: Polyrepo (rejected — coordination overhead), monolith (rejected — scaling concerns)
- Confidence: 78%
- Unknowns: CI performance at 20+ packages, developer onboarding friction
Time: 30 seconds.
Total Time: Under 4 Minutes¶
Compare that to a 45-minute architecture meeting where half the team is checked out and the decision is made by whoever talks loudest.
When to Use This¶
- Technology selection (framework, database, hosting)
- Architecture patterns (monolith vs micro, REST vs GraphQL)
- Migration decisions (rewrite vs refactor vs abandon)
- Any decision you'll need to justify in 6 months
Revisiting Decisions¶
Six months later, someone asks "why monorepo?"
Full context. Full reasoning. Full list of alternatives you already considered. No archaeology required.
Tips¶
Skip CONCLAVE for small decisions
Not everything needs 9 voices. For smaller calls, go straight to MIRROR → ECHO. Save CONCLAVE for the decisions that keep you up at night.
Record the unknowns
ECHO captures unknowns for a reason. Revisit them periodically. When an unknown becomes known, update the echo.
Confidence thresholds
Below 60% confidence? The decision isn't ready. Research more (RECON), debate more (CONCLAVE), or accept you're making a bet and document it honestly.