Dynamic context pruning patterns (OpenCode DCP / Sleev). Meta has native auto-compact; use these rules to manage long sessions.
Scanned 9/2/2026
Install to Claude Code
npx -y skills add nuroctane/nur-cli --skill context-pruning --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Context Pruning?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/nuroctane-context-pruning)More formats (shields.io, HTML) on the badges page.
---
name: context-pruning
description: "Dynamic context pruning patterns (OpenCode DCP / Sleev). Meta has native auto-compact; use these rules to manage long sessions."
---
# Context pruning (DCP-inspired)
Upstream: https://github.com/Opencode-DCP/opencode-dynamic-context-pruning
Successor focus: https://sleev.ai (`npm i -g sleev`)
OpenCode's DCP plugin is **OpenCode-specific**. Meta implements the same goals natively:
## Meta native behavior
- Auto-compact when context pressure is high (~55% of window, once per turn)
- Manual `/compact` slash command
- Tool results are capped; prefer re-query over replaying huge dumps
## Practices for long sessions
1. After a milestone, summarize and drop raw tool blobs (user can `/compact`)
2. Prefer graphify/plur recall over re-grepping the whole repo
3. Don't re-read files already summarized unless editing
4. Parallel read-only tools only — mutating tools stay sequential
5. If using OpenCode elsewhere: `opencode plugin @tarquinen/opencode-dcp@latest --global`
## Compress modes (conceptual)
- **range** — compress a span of turns into one summary
- **dedupe** — identical tool+args keep latest output only
- **purge errors** — drop large error inputs after N turns
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!
Ultra-compressed communication mode. Cuts token usage ~75% by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.