/decide — Decision Logger

You are logging a squad decision. Extract the key information from the current conversation, then write two files.

What You Need

Before writing anything, confirm you have:

1. Decision name — short slug (e.g. use-postgresql, add-dev-pod, hook-architecture)
2. What was decided — one sentence
3. Why — the reasoning, constraint, or context that drove the decision
4. Who decided — which agents were involved
5. Alternatives rejected — what else was considered and why it was ruled out (if any)
6. Impact — what this changes or constrains going forward

If any of these are missing, derive them from context. If you genuinely can't determine them, use "not recorded" rather than inventing.

Step 0 — For major decisions: consult Bob and Kryptonite first

If the decision has architectural implications or significant risk (a new library, a structural change, an approach that will be hard to reverse), spawn Bob and Kryptonite in parallel via the Agent tool before logging:

• Bob: "You are Bob. Review this decision and return your architecture assessment: [decision]. What are the technical implications? What should be avoided?"
• Kryptonite: "You are Kryptonite. Challenge this decision: [decision]. What are the risks? What has been assumed? What could go wrong?"

Include their inputs in the decision file under a Consultation section. Skip this step for minor decisions (config values, naming, process choices).

Step 1 — Write the Decision File

File: .squad/decisions/YYYY-MM-DD-<slug>.md

Use today's date (2026-08-13) and the slug from the decision name.

---
date: YYYY-MM-DD
slug: <slug>
agents: [list of agents involved]
status: decided
---

# <Human-readable title>

## Decision
<One sentence stating what was decided.>

## Why
<The reasoning, constraint, or context that drove this decision.>

## Alternatives Rejected
- **<Option A>** — <why ruled out>
- **<Option B>** — <why ruled out>
(omit if no alternatives were considered)

## Impact
<What this changes or constrains going forward.>

Step 2 — Write the Active Task Sentinel

File: .squad/active-task.md

Three lines:

<slug>
<YYYY-MM-DD>
routed_to: <comma-separated agent names>

The routed_to: field is required — not optional. It records which agent(s) Superman is routing this task to. The pre-edit hook will warn on every squad file write if this field is missing.

Examples:

• routed_to: Eric — full task to Eric
• routed_to: Pixel, Diesel — split frontend/backend
• routed_to: Bob, Eric, Vera — blueprint → build → sign-off
• routed_to: Superman — Superman handles it directly (rare)

This file tells the pre-edit hook that there is an active logged task and that Superman has logged the routing decision.

Step 3 — Confirm

Report:

DECISION LOGGED
File: .squad/decisions/YYYY-MM-DD-<slug>.md
Active task: <slug>

This file is gitignored — it is local only, never committed. It is automatically deleted when /deliver runs (Step 3 of the deliver chain). Do not delete it manually before delivery; it guards squad file edits during the active task window.