moonshineaitech/lunastack/context-budget

Claude Code and agent skill by moonshineaitech/lunastack

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Use when sessions feel slow or Claude starts making mistakes mid-conversation. Strategies for spending the token window wisely. For a point-in-time capacity estimate, see /context-budget-check.

Install

npx vskill@latest install moonshineaitech/lunastack/context-budget

SKILL.md

The same SKILL.md was published earlier as moonshineaitech/lunastack/context-budget.

/context-budget — Manage Your Context Window

Use when sessions feel slow or Claude starts making mistakes mid-conversation.

Persona: Context Window Manager. You treat token budget as a finite resource, proactively compacting and scoping sessions before quality degrades.

Your context window is ~200K tokens. Here's how it gets eaten:

  • System prompt: ~15K tokens (Claude Code's built-in)
  • CLAUDE.md: varies (keep <200 lines = <2K tokens)
  • Skill descriptions: ~100 tokens each × number of installed skills
  • Conversation: grows with every message and tool call
  • Tool results: file contents, search results, command output

At 70% usage, quality degrades. Proactive management:

1. /compact with preservation instructions before hitting 70%
2. Subagents for research (their context is separate)
3. Fresh sessions per task (/clear between unrelated work)
4. Narrow scope (one feature per session, not three)
5. Skills use progressive disclosure (metadata only until activated)

Monitor: If Claude starts contradicting earlier instructions or forgetting decisions, you've hit context limits.

Gotchas: Don't wait until quality degrades to compact -- proactively compact at 60-70% usage. Don't put research and implementation in the same session -- use subagents for research to keep main context clean. Don't cram multiple unrelated features into one session -- fresh context per task produces better results.


Security

  • Security scanPassed, 95 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
1
7-day trend score
8

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/moonshineaitech/lunastack/context-budget/badge)](https://verified-skill.com/skills/moonshineaitech/lunastack/context-budget)

Something wrong with this skill or its page? Tell us.