Decision Logger

Decision Logger

Capture the reasoning behind key product decisions as structured, append-only traces. This builds the "event clock" — the record of why things became true, not just what is true now.

When to Use

• After a PRD is approved or rejected
• After a scope change or pivot
• After a launch/kill/iterate decision on an experiment
• After a prioritization shift (quarterly planning, reprioritization)
• After a significant stakeholder alignment (exec review, cross-functional agreement)
• When asked "log this decision", "record why we decided this", or "decision trace"

Process

1. Identify the Decision

Ask if not obvious from context:

• What was decided? (1 sentence)
• What triggered this decision? (signal, meeting, data, escalation)

2. Capture the Trace

Gather the following — from conversation context, user input, or by asking:

• Decision: What was decided (1 sentence)
• Date: When the decision was made
• Context: What prompted this — the signal, conversation, or data point
• Alternatives considered: What other options were on the table (at least 2)
• Evidence: What data, insight, or customer signal drove the call
• Stakeholders: Who was involved, who approved, who dissented
• Expected outcome: What we expect to happen as a result
• Risk / reversal cost: How hard is this to undo if wrong
• Tags: strategic-pillar, metric, segment, experiment (for retrieval)

3. Write the Entry

Save to memory/decisions/YYYY-MM-DD_short-slug.md using this format:

# [Decision title]

**Date:** YYYY-MM-DD
**Type:** [PRD approval | Scope change | Launch/Kill | Prioritization | Alignment | Other]
**Tags:** [pillar, metric, segment]

## Decision
[1-2 sentences: what was decided]

## Context
[What prompted this — the trigger]

## Alternatives Considered
1. **[Option A]** — [brief description, why not chosen]
2. **[Option B]** — [brief description, why not chosen]

## Evidence
- [Data point, customer insight, or signal that drove the call]

## Stakeholders
- **Decision maker:** [who]
- **Consulted:** [who]
- **Informed:** [who]

## Expected Outcome
[What we expect to happen, and how we'll know]

## Risk / Reversal Cost
[How hard is this to undo — low/medium/high, and why]

4. Confirm with User

Show the drafted entry. Ask the user to confirm or adjust before saving.

Output

A timestamped decision trace saved to memory/decisions/. Append-only — never overwrite existing entries.

Over time, these traces enable:

• Recall: "Why did we decide X?"
• Pattern detection: "We've killed 3 pricing experiments in a row"
• Drift detection: "Strategy says X but decisions are moving toward Y"
• Counterfactual reasoning: "What if we'd chosen the alternative?"

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: hardiktiwari
• Source: hardiktiwari/PM-operating-OS
• License: MIT

Install and usage instructions live in the source repository linked above.