Agent work receipt

An agent work receipt is a short note you fill out after an agent run. It helps a reviewer see what was requested, what files were read or changed, what evidence supports completion, and what still needs human review.

It is not a guarantee that the agent was correct. It is a review aid.

Use this after running the Steven Workflow Demo with playground/FIRST_PROMPT.md.

What to check after the playground run

After you run playground/FIRST_PROMPT.md, check the workspace before you judge the answer:

1. A new draft exists under playground/Blog/Published/.
2. A new learning note exists under playground/wiki/outputs/agent-learnings/.
3. The draft was based on playground/raw/thoughts/2026-04-23-why-most-ai-feels-generic.md.
4. The run used playground/wiki/_index.md, prior published work, summaries, and writing rules before drafting.
5. The draft explains how it differs from the prior article playground/Blog/Published/2026-04-10-ai-workflow-is-the-product.md.
6. The learning note records a reusable writing rule or skip condition for the next run.
7. git diff or your editor shows only the files you expected the agent to create or edit.

If the agent only returns a chat answer and writes no files, the playground run did not complete the workflow.

Copyable template

# Agent Work Receipt

## Run

- Date:
- Runner:
- Prompt used:
- Workspace:

## Request

What did the user ask the agent to do?

## Files The Agent Said It Read

| File | Why it mattered |
|------|-----------------|
|  |  |

## Files Created Or Changed

| File | Created or changed? | What changed |
|------|---------------------|--------------|
|  |  |  |

## Evidence To Check

- [ ] A new draft exists under `playground/Blog/Published/`.
- [ ] A new learning note exists under `playground/wiki/outputs/agent-learnings/`.
- [ ] The draft uses the source file named in the prompt.
- [ ] The draft checks against prior published work instead of repeating it.
- [ ] The learning note is reusable for a later run.
- [ ] The diff contains only expected files.

## Still Needs Human Review

- [ ] The draft is factually correct.
- [ ] The draft is not overclaiming what the repo or demo can do.
- [ ] The writing quality is good enough to publish or continue editing.
- [ ] The learning note is specific enough to help the next run.

## Honest Boundary

This receipt makes the work easier to inspect. It does not prove that the agent is correct, complete, or safe to publish without review.

Filled example: playground first prompt

This is an example receipt for a successful local run of playground/FIRST_PROMPT.md. It uses only public repo files. It does not claim that a hosted runtime exists.

The exact output filenames may differ in your run. The important check is that the agent writes a new draft under playground/Blog/Published/ and a new learning note under playground/wiki/outputs/agent-learnings/.

Run

• Date: 2026-05-10
• Runner: Local coding agent or AI IDE with repository file access
• Prompt used: playground/FIRST_PROMPT.md
• Workspace: playground/

Request

Turn playground/raw/thoughts/2026-04-23-why-most-ai-feels-generic.md into a publishable blog draft. Before writing, read the local instructions, wiki index, prior published work, and learning notes. After drafting, write a reusable learning note.

Files The Agent Said It Read

File	Why it mattered
playground/AGENTS.md	Defines the playground workspace rules and write-back expectation.
playground/FIRST_PROMPT.md	Contains the user request and success condition.
playground/raw/thoughts/2026-04-23-why-most-ai-feels-generic.md	Provides the raw thought material for the draft.
playground/wiki/_index.md	Points the agent to the relevant local context.
playground/Blog/Published/2026-04-10-ai-workflow-is-the-product.md	Lets the agent check whether the topic repeats prior work.
playground/wiki/summaries/2026-04-12-agent-products.md	Supplies reusable language and core judgment.
playground/wiki/outputs/agent-learnings/2026-04-15-writing-rules.md	Preserves prior writing rules for this workspace.
capabilities/content-creation.md	Gives the article workflow and output contract.

Files Created Or Changed

File	Created or changed?	What changed
playground/Blog/Published/2026-04-23-why-most-ai-feels-generic.md	Created by the run	Blog draft based on the raw thought file.
playground/wiki/outputs/agent-learnings/2026-04-23-writing-from-generic-ai-note.md	Created by the run	Reusable rule about turning raw anti-generic-AI notes into a sharper article.

Evidence To Check

• [ ] git status --short playground/Blog/Published playground/wiki/outputs/agent-learnings shows one new draft and one new learning note.
• [ ] The new draft names or clearly uses playground/raw/thoughts/2026-04-23-why-most-ai-feels-generic.md as its source.
• [ ] The new draft does not simply repeat playground/Blog/Published/2026-04-10-ai-workflow-is-the-product.md.
• [ ] The new learning note is a reusable rule, not a one-line summary of the finished article.
• [ ] The run did not edit private notes, secrets, unpublished vault files, or unrelated directories.

Still Needs Human Review

• [ ] Whether the draft is accurate and worth publishing.
• [ ] Whether the distinction from the prior article is strong enough.
• [ ] Whether the new learning note should stay as written, be merged into an existing rule, or be deleted.

Honest Boundary

This receipt makes the run auditable by pointing to files and checks. It does not prove that the agent made the right editorial judgment.