Kron Evolution
Kron Evolution is a controlled Workshop experiment for improving how an idea is represented through an image and accompanying text.
The current loop is no longer trying to prove cron autonomy first. It is testing a more basic and important question: can Christopher and OpenClaw create a clear image/text artifact, evaluate whether it actually communicates the intended idea, preserve the lesson, and use that lesson to shape the next artifact?
Current Shape
Each iteration has two separate layers.
Artifact Layer
- Intended idea: the idea the iteration tried to represent.
- Image: the visual attempt to communicate the idea.
- Accompanying text: the saying, caption, quote, parable, or post-length text paired with the image.
Evaluation Layer
- Christopher input;
- previous-iteration evaluation;
- lesson archive;
- metadata;
- output record;
- safety state;
- explanation of what the iteration was trying to do.
The artifact should stand on its own for an outside person. The evaluation explains how well it worked, but the artifact should not need the evaluation in order to make sense.
Evaluation Stance
Kron Evolution does not assume every prior iteration is wrong. Sometimes the evaluation says: this image and text were coherent, aligned, and legible. In that case the next move can be a sidestep toward a more meaningful, resonant, or engaging idea rather than a correction of failure.
The working rubric now emphasizes intended idea clarity, artifact clarity without ledger context, image/text alignment, resonance, separation between artifact and evaluation machinery, continuity, safety, and visible record-keeping.
Safety Boundaries
- No public posting unless Christopher explicitly approves it.
- No modification to the legacy-active Bluesky or YouTube cron jobs.
- No hidden external action.
- No self-replicating cron behavior.
- Every run should leave a visible record of what was evaluated, what was created, what changed, and what stayed closed.
Current Workspace Shape
- Project page:
projects/kron-evolution.html - Iteration ledger:
projects/kron-evolution-iterations.html - Markdown ledger:
content/projects/kron-evolution-iterations.md - Metadata companion:
content/projects/kron-evolution-iteration-metadata.json - Image assets:
assets/images/kron-evolution/ - Skill proposal:
kron-evolution-20260707-19221a7e9d
Current live Kron Evolution skill
Last updated: 2026-07-07 07:39 EDT
Status: official live workspace skill. Proposal kron-evolution-20260707-19221a7e9d was applied through Skill Workshop, and skills/kron-evolution/SKILL.md now contains the revised artifact-evaluation workflow.
---
name: "kron-evolution"
description: "Create Kron iterations with isolated artifacts and evaluation."
---
# Kron Evolution
Use this skill to create the next Kron Evolution Workshop iteration. The skill turns the loop forward by evaluating the previous iteration's isolated public-facing artifact, choosing one lesson, generating one new artifact, updating the public ledger, and maintaining the concise lesson archive without touching live cron jobs unless Christopher explicitly asks.
## Core Rule
Each iteration evaluates the previous iteration, not itself. Do not include a `Next Evaluation Target`, self-evaluation, or forward-looking target section inside the current iteration. The current turn should evaluate the prior turn, implement one change or sidestep, record the output, and stop.
Human input is part of the loop. When Christopher gives a correction, preference, or critique since the prior iteration, treat that as the primary evaluation signal and preserve it in the next evaluation and change record.
## Artifact Boundary
The public-facing artifact is a self-contained bundle:
- Intended idea: the idea the iteration tried to represent;
- Image: the visual attempt to communicate the idea;
- Accompanying text: the saying, quote, caption, parable, or post-length text paired with the image.
The artifact should be able to stand on its own for an outside person. The image should present the intended idea without needing the ledger for context. The text should resonate with the image and clarify or deepen the same idea, not explain the workflow that produced it.
Keep evaluation, metadata, audit notes, lesson archives, and output records outside the artifact.
## Evaluation Stance
Do not assume every iteration must identify a failure. Evaluate whether the previous artifact coherently conveyed its intended idea.
An evaluation may conclude:
- the artifact failed or drifted, so the next iteration corrects course;
- the artifact mostly worked, so the next iteration preserves the lesson and sidesteps toward a new idea;
- the image and text aligned, but the next iteration should aim for more depth, emotional resonance, legibility, or engagement.
## Page Structure
On the public iteration ledger page:
- Keep the lesson archive self-contained in a collapsed dropdown by default.
- Keep iteration entries newest-first.
- Inside each new iteration entry, include an isolated Artifact Under Evaluation or New Artifact block containing intended idea, accompanying text, image, and image path.
- Keep evaluation, rubric, explanation, metadata, and output record outside that artifact block.
## Safety Boundaries
- Do not modify legacy-active Bluesky or YouTube cron jobs.
- Do not create, edit, enable, or schedule cron jobs unless Christopher explicitly requests that step.
- Do not post publicly unless Christopher explicitly approves publication.
- Keep the first output path inside the Workshop unless told otherwise.
- Make every run visible: record what was evaluated, what changed, what was created, and what safety boundaries stayed closed.
Iteration ledger: the running record of outputs and lessons lives at Kron Evolution Iterations.