Review Prompt

# Weekly Gap Snapshot Review Prompt

Run this prompt against your AI once a week. It reads all files in `gap_snapshots/` and surfaces patterns.

---

**Prompt:**

> Read every file in `gap_snapshots/` from the last 30 days. For each snapshot, note:
> 1. The question or decision that triggered it
> 2. What was missing
> 3. Whether the outcome has been filled in (check `outcome_filled_in_later` field)
>
> Then produce a report with three sections:
>
> **1. Recurring knowledge holes.** Which topics or question types keep triggering gap snapshots? These are your system's blind spots.
>
> **2. Unfilled outcomes.** Which snapshots are older than 14 days with `outcome_filled_in_later` still null? These should either be filled in or closed as "never resolved."
>
> **3. Highest-value closures.** For the recurring holes, suggest one specific action to close each: "add X to atlas/", "conduct Y interview," "build Z data source."
>
> Save the report to `gap_reports/YYYY-MM-DD.md`.

---

## What to Do With the Report

- **Recurring holes** → close the highest-frequency ones first. Add atlas files, gather data, or explicitly decide the hole stays open.
- **Unfilled outcomes** → backfill the `outcome_filled_in_later` field with what actually happened. If the decision was punted indefinitely, mark it as "deferred" and close the snapshot.
- **Highest-value closures** → add to `current-priorities.md` or an appropriate workstream.

Over time, gap snapshots become your highest-quality training data — real examples of where your system fell short, labeled with what would have helped.