---
name: aiidiots-field-notes
description: Capture reusable lessons after meaningful agent work. Use when the user says "AII field note", "what did we learn", "install field notes", or asks the agent to turn work into a self-learning loop.
---

# AIIdiots Field Notes

Use this skill after meaningful work, especially if the run exposed a reusable
failure mode, setup gotcha, tactic, privacy rule, routing lesson, or product
idea.

Canonical invocation:

```text
AII field note: what did we learn from this?
```

This is different from `AII first`.

- `AII first: <task>` applies doctrine before work.
- `AII field note: ...` captures reusable learning after or between work.

## Field Note Pass

Return a field note with:

- task/context
- what happened
- what changed
- success or failure mode
- reusable lesson
- evidence or receipt paths
- privacy class: `private`, `sanitized-public-candidate`, or `public`
- proposed doctrine update
- follow-up work

## Privacy Rules

- Default raw notes to private.
- Do not publish raw private context, transcripts, secrets, customer/business
  data, account details, or sensitive local paths.
- Separate the private field note from the public lesson.
- Publish only sanitized lessons, rules, receipts, and examples after review.
- If the lesson came from a person, workplace, or private project, preserve the
  operating lesson and remove the identifying details.

## Output Shape

```markdown
# AII Field Note

## Context
What task or run produced this lesson?

## What Happened
What did the agent/user/system do?

## Lesson
What should a future agent do differently?

## Evidence
Receipts, tests, files, commands, or links that support the lesson.

## Privacy Class
private | sanitized-public-candidate | public

## Doctrine Update
The rule, route brief, checklist, or task pack this should update.

## Follow-Up
What should happen next?
```

## Agent Rule

Do not let useful learning disappear into a transcript. If the work taught a
future agent how to behave better, file the field note before claiming the
task is fully done.
