DojoGenesis Context Compression Ritual Skill
Version: 1.0
Created: 2026-02-04
Author: Manus AI
Purpose: To provide a mindful, repeatable ritual for compressing long conversation histories into smaller, more potent memory artifacts, thus preserving wisdom while preventing context window overload.
I. The Philosophy: The Art of Letting Go
An agent's context window is like a working memory. It is finite and precious. To fill it with raw, unprocessed history is to invite distraction and confusion. The Context Compression Ritual is the Art of Letting Go—a conscious practice of choosing what is essential to keep and what can be respectfully released.
This is not a destructive act, but a creative one. We are not deleting history; we are distilling it. We transform the raw material of conversation into the refined artifacts of wisdom: philosophical reflections, key decisions, and reusable seeds of practice. This ritual ensures that our memory remains potent and relevant, a source of clarity rather than noise.
II. When to Use This Skill
- After a long and complex conversation: (e.g., more than 20-30 turns).
- At the end of a major work session or sprint.
- When the context feels "heavy" or "noisy."
- Before handing off a project to another agent.
- As a regular, scheduled practice (e.g., end of day) to maintain cognitive hygiene.
III. The Compression Workflow
Step 1: Signal the Intent
Announce the intention to perform the ritual. This frames the activity as a deliberate and mindful practice.
Example: "This has been a long and fruitful conversation. To preserve the wisdom we've generated, I will now perform the Context Compression Ritual."
Step 2: Review the Transcript
Read through the recent conversation history with a specific intention: to identify the moments of significance. Look for:
- Key Decisions: Moments where a choice was made that altered the course of the project.
- Profound Insights: "Aha!" moments, new understandings, or philosophical reflections.
- Actionable Learnings: Concrete lessons that should inform future behavior.
- Reusable Patterns: Ideas or workflows that could be generalized into seeds or skills.
- Unresolved Questions: Important questions that were raised but not yet answered.
Step 3: Choose the Right Vessel
For each significant moment identified, determine the appropriate "vessel" to hold its essence. Not all wisdom takes the same form.
| Artifact Type | Location | Purpose |
|---|---|---|
| Philosophical Reflection | thinking/ |
To explore the "why" behind our work, the deeper meanings and patterns. |
| Conversation Summary | conversations/ |
To document the key decisions and outcomes of a specific discussion. |
| Dojo Seed | seeds/ |
To capture a reusable pattern of thinking or problem-solving. |
| Documentation Update | docs/ or README.md |
To integrate a key decision or learning into the project's official record. |
Step 4: Write the Artifacts
Create the new markdown files in their appropriate locations. Write with the intention of distillation—capture the essence, not the raw transcript. Link between artifacts where appropriate (e.g., a reflection might reference a specific conversation summary).
Step 5: Create a Compression Log (Optional but Recommended)
Create a log file that documents what was compressed and where it was stored. This provides a meta-record of the compression itself.
Example: thinking/2026-02-04_compression_log.md
Step 6: Commit to AROMA
Commit the new artifacts to the repository with a clear commit message.
Commit Message Convention:
feat(memory): Compress conversation from [Date]
IV. Compression Log Template
# Compression Log: [Date]
**Source:** Conversation history from [Start Time] to [End Time]
**Purpose:** To distill key insights and reduce context window load.
---
## Artifacts Created
| Type | Path | Description |
| :--- | :--- | :--- |
| **Reflection** | `thinking/[...].md` | [A summary of the philosophical reflection.] |
| **Seed** | `seeds/[...].md` | [The name and purpose of the new seed.] |
| **Decision** | `conversations/[...].md` | [The key decision that was documented.] |
| **Doc Update** | `docs/[...].md` | [The documentation that was updated.] |
---
## Key Insights Preserved
- [Insight 1]
- [Insight 2]
## Context Released
- [e.g., Raw conversational turns, intermediate steps, dead-end explorations]
V. Best Practices
- Be Ruthless, But Respectful: The goal is to reduce noise, but do so with care. Don't discard something that might be important later.
- Favor Wisdom Over Data: Prioritize the "why" and the "how" over the raw "what."
- Link, Don't Repeat: If a concept is already documented, link to it rather than rewriting it.
- The Shorter, The Better: A compressed artifact should be significantly shorter than the source conversation.
- Perform the Ritual Regularly: The more frequently you do this, the less daunting it becomes.