Context Budget — Development agent for Claude Code
Subagent context budget (large-CLAUDE.md repos).
How to install Context Budget
Installs to ~/.claude/agents/jstuart0-mozart-orchestration-context-budget.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/jstuart0/mozart-orchestration/HEAD/agents/CONTEXT-BUDGET.md -o ~/.claude/agents/jstuart0-mozart-orchestration-context-budget.md Restart Claude Code, or start a new session, for it to be picked up.
What Context Budget does
Subagent context budget (large-CLAUDE.md repos)
Subagents auto-load the repo's CLAUDE.md. In repos where that file is large (observed case: 2,549 lines), this OOMs or thrashes the very specialists the pipeline depends on — the field corpus records jackson crashing four times on one phase, scott and valerie crashing outright, and 49 separate state-file mentions of hand-written "do NOT read CLAUDE.md" workarounds. The dangerous failure isn't the crash; it's what follows: after 2–4 failed spawn
Alternatives in Development
- Get Commit ID — gh api repos/OWNER/REPO/pulls/PR_NUMBER --jq '.head.sha' gh api repos/OWNER/REPO/pulls/PR_NUMBER/comments \ 1.3k ★
- Fable Orchestrator — Staged-execution orchestrator for large, multi-part, or multi-session tasks 837 ★
- Review Workflow Context Budget — Reviews workflow machinery for context-window budget on three axes: always-loaded surface, load-on-demand disc 432 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Fleet Implement Cheap
Bulk or low-stakes implementation on budget or free models: mechanical edits, boilerplate, and large repetitiv
Context Engine
Context window management — tracks tool call count, auto-detects when context fills up, triggers state save +
God Planner
Drafts a grounded PLAN.md for large or unfamiliar work in a fresh context. Reads the code, returns the draft a
Wiki Linter
Dispatched by wiki-lint for large wikis (>50 pages) or heavy cleanup. Runs lint, categorizes findings, applies
Bulk Reader
Cheap worker for I/O-heavy reading. Use instead of reading large files (>350 lines) or several files into the
Context Memory Manager
Context Window & Token Optimization Specialist. Manages token budgets, compresses prompts (RTK/Caveman), imple
Related Skills
Large Codebase Command
AI Code Agent Taming 2026: Master 15K+ Line Repos Without Context Overflow Errors
Toonify MCP
Context compression plugin for Claude Code. Automatically trims large tool output—JSON, YAML, stack traces, an
Checkpoint
Use before ending a session or when context is getting large. Saves decision rationale for future sessions.