Share your Fluently 4D cycles with the community
Use your AI agent + GitHub token to open a PR on your behalf. Works for individuals and teams sharing a GitHub PAT.
Fork the repository and clone it locally to your machine.
git clone https://github.com/YOUR-USERNAME/fluently.git
cd fluently
Create a descriptive branch for your new Fluently 4D cycle.
git checkout -b knowledge/add-your-topic
Create a new .yaml file in /knowledge/ with all four dimensions fully explained.
touch knowledge/your-topic-name.yaml
Commit your changes with a descriptive message following the format.
git add knowledge/your-topic.yaml
git commit -m "docs: add Fluently 4D cycle for [topic]"
git push origin knowledge/add-your-topic
Go to GitHub and open a PR. Include a description of what your Fluently 4D cycle covers.
Open PR on GitHub →idUnique identifier in kebab-case (no spaces, lowercase, hyphens only).
id: bug-fix-prioritization
titleClear, actionable title (5-100 characters).
title: "Bug Fix Prioritization"
domain
One of: coding, writing, research, customer-support, education, legal, healthcare, general
domain: coding
score_hintsRelative weights summing to 1.0 across 4 dimensions.
score_hints:
delegation: 0.25
description: 0.25
discernment: 0.25
diligence: 0.25
Describe whether this task should be:
How should users frame their request to the AI?
How do you know if the AI's output is trustworthy?
What human accountability measures apply?
id: your-unique-id
title: "Clear, Actionable Title"
domain: coding # coding, writing, research, customer-support, education, legal, healthcare, general
dimensions:
delegation:
description: How should delegation/augmentation of this task work?
example: AI suggests options, human approves before acting.
antipattern: Fully automating without any human checkpoint.
description:
description: What context/framing makes the AI most useful here?
example: Include repo context, examples of desired output, and constraints.
antipattern: Vague or missing context leads to irrelevant suggestions.
discernment:
description: How do you evaluate if the AI output is trustworthy?
example: Cross-check AI findings against test results and peer judgment.
antipattern: Accepting AI output without independent verification.
diligence:
description: What human accountability is required after AI involvement?
example: Lead engineer signs off before the output is acted on.
antipattern: No approval process or audit trail.
score_hints:
delegation: 0.25 # Sum must equal 1.0
description: 0.25
discernment: 0.25
diligence: 0.25
# The collaboration block is required. It captures how the 4Ds sequence as
# human-AI conversation clusters — each D is a chain of related prompts, not a single prompt.
# pattern: linear | linear_with_loops | cyclic | iterative | branching
collaboration:
pattern: linear_with_loops
description: "One-line description of how Ds flow for this task."
sequence:
- step: 1
d: delegation
label: "Negotiate AI scope and autonomy"
example_prompts:
- speaker: human
text: "Can you handle X automatically and flag Y for me?"
- speaker: ai
text: "I can flag Y with confidence levels — want me to auto-handle only Z?"
- speaker: human
text: "Yes — auto-handle Z, surface everything else."
triggers_next: "Autonomy boundaries agreed."
- step: 2
d: description
label: "Provide context and constraints"
example_prompts:
- speaker: human
text: "Here is the context, constraints, and examples."
- speaker: ai
text: "Should I prioritize A or B?"
- speaker: human
text: "A first, then B."
triggers_next: "AI has sufficient context."
- step: 3
d: discernment
label: "Evaluate AI output"
example_prompts:
- speaker: human
text: "Item 3 looks like a false positive — is it?"
- speaker: ai
text: "Possibly — given X, this could be dismissed."
- speaker: human
text: "Agreed, dismiss it."
triggers_next: "Output validated."
loop_back:
to: delegation
condition: "If quality is consistently poor."
reason: "Scope or autonomy level needs renegotiation."
- step: 4
d: diligence
label: "Approve and document"
example_prompts:
- speaker: human
text: "Approving and logging the decisions."
triggers_next: "Cycle complete. Restarts for next instance."
can_restart: true
transitions:
- from: delegation
to: description
trigger: "Scope agreed."
- from: description
to: discernment
trigger: "AI delivers output."
- from: discernment
to: diligence
trigger: "Output validated."
- from: discernment
to: delegation
trigger: "Quality too low — re-scope."
is_loop_back: true
- from: diligence
to: delegation
trigger: "Next instance — restart."
is_cycle_restart: true
tags:
- your-topic
- category
- application
contributor: "Your Name or GitHub Handle"
version: "1.0.0"Each dimension should be 100-300 words. Be specific and practical.
The CI will automatically validate your YAML against the schema when you open a PR. Fix any errors before submitting.
Use this template when opening your pull request:
## What's the Fluently 4D Cycle About?
Brief explanation of your cycle and why it matters.
## When Should This Be Used?
Real-world scenarios where this guidance helps.
## Verification Checklist
- [x] Schema validation passes locally (`node scripts/validate-knowledge.js`)
- [x] All 4 dimensions are complete (description, example, antipattern)
- [x] score_hints sum to 1.0
- [x] score_hints reflect relative importance
- [x] collaboration block is present with pattern, sequence (≥2 steps), and transitions
- [x] example_prompts show realistic human↔AI exchanges
- [x] YAML syntax is valid (checked in editor)Need help or have ideas? Reach out to the community on GitHub.