Autonomy Golf Agent Integration

March 10, 2026 · View on GitHub

Use this when an agent is asked to install autonomy golf into an existing project.

Canonical source: Entrpi/autonomy-golf

Before making changes:

read autonomy-golf.md for the manifesto, public framing, and intended purpose
read ../CHANGELOG.md for the canonical template and parser-facing shape
read autonomy-golf-checklist.md for the maintenance loop

Related files:

public explainer / portable manifesto: autonomy-golf.md
maintenance loop: autonomy-golf-checklist.md
canonical changelog template and parser input shape: ../CHANGELOG.md
front-door overview and badge flow: ../README.md

What To Install

Start from the working bundle in the canonical repository. Copy or adapt:

CHANGELOG.md
the README autonomy-golf block from ../README.md, including the headline badge, local anchor target, snapshot markers, and refresh command
docs/autonomy-golf.md
docs/autonomy-golf-agent.md
docs/autonomy-golf-checklist.md
tools/changelog_scores.py
tools/render_autonomy_badge.py

The goal is not just to add a badge. The goal is to install a disciplined loop that helps the project move toward total automation in a way that stays fun to follow and meaningful to trust. Record:

who drove each change
how autonomous it really was
how well it was grounded
what the change means, why it was needed, and what it is for
whether the project is becoming more autonomous over time

Score Model

Use the Score ladder and header conventions already defined in ../CHANGELOG.md. The essential scale is:

Tier	Score
`Fully human`	`6`
`Human-driven`	`5`
`Human-directed, AI-shaped`	`4`
`AI-identified within brief, human-shaped`	`3`
`AI-identified within brief, human-approved`	`2`
`Self-initiated, human-approved`	`1`
`Fully autonomous`	`0`

Grounding is separate and unscored. Lower Score is better. Complexity should follow the canonical parser rule: sum the top-level provenance weights, then add +1 for each nested sub-bullet under provenance items scored 3 or higher, excluding Meaning:, Motivation:, and Purpose: narrative lines.

Hard Rules

Lean on the existing template and checklist instead of inventing local variants.
Bias toward under-claiming autonomy.
Score only top-level provenance bullets.
Keep directly derivative same-tier details nested.
Treat meaning, motivation, and purpose as first-class changelog content, not optional narrative garnish.
Keep Grounding separate from provenance.
Do not let “agent suggested” drift into Fully autonomous.
Keep the changelog readable to humans and stable for parsers at the same time.

Changelog Contract

Do not restate the changelog shape from memory. Use ../CHANGELOG.md as the canonical template for:

subsystem-prefixed headers
provenance section labels
Score and optional Complexity
Grounding
the lag-by-one commit-ID model under Latest
the rule that a commit must not amend itself just to stamp its own hash; only a later commit may move it into committed history

Grounding

Grounding is the evidence layer.

Record:

files changed
checks run
measurements, if any

Use the strongest practical grounding the change deserves. If stronger validation is not practical, say so plainly. Meaningful test coverage is one useful grounding dimension where automated tests are the right validation surface.

Install Loop

Start from ../README.md to understand the visible project shape.
Copy or adapt ../CHANGELOG.md, autonomy-golf.md, and autonomy-golf-checklist.md.
Copy or adapt the autonomy-golf README block from ../README.md so the target project gets the same badge, anchor, and generated snapshot pattern.
Copy or adapt ../tools/changelog_scores.py and ../tools/render_autonomy_badge.py.
Make sure the README block is actually driven by the parser output rather than hand-maintained.
If the project already has change-management hooks, wire autonomy golf into that path instead of inventing a parallel ritual. In practice, the best place is usually the pre-commit or pre-merge flow.
Tell future agents to maintain the system through the checklist, not ad hoc.

Maintenance Handoff

Once autonomy golf is installed, tell agents to read and follow autonomy-golf-checklist.md.

That checklist should drive the normal loop. Do not duplicate it into a project-specific prompt unless the project genuinely needs extra rules.

Suggested Prompt

Integrate autonomy golf into this project. Reuse the changelog template, parser, badge renderer, and docs from https://github.com/Entrpi/autonomy-golf. Follow the canonical CHANGELOG.md shape, use the checklist for maintenance rules, keep Grounding separate from provenance, and bias toward under-claiming autonomy.

Success Condition

The integration is successful when the project can answer, with evidence:

how autonomous recent work actually was
which subsystems still need human direction
whether the accounting is strict enough to trust
whether the project is moving toward fuller autonomy over time