Agent Skills › AjayIrkal23/agentic-mercy-10x › lean-ctx

lean-ctx

GitHub

通过81个MCP工具和多种读取模式，对代码、Shell输出进行高压缩比处理，大幅减少LLM上下文消耗。支持文件读取、搜索、编辑及架构分析，适用于代码探索与开发场景。

skills/lean-ctx/SKILL.md AjayIrkal23/agentic-mercy-10x

触发场景

需要高效读取大文件或大量代码时执行Shell命令并需压缩输出结果时进行代码搜索或目录结构浏览时需要进行锚点式文件编辑以避免冗余文本时

安装

npx skills add AjayIrkal23/agentic-mercy-10x --skill lean-ctx -g -y

更多选项

不安装直接使用

npx skills use AjayIrkal23/agentic-mercy-10x@lean-ctx

指定 Agent (Claude Code)

npx skills add AjayIrkal23/agentic-mercy-10x --skill lean-ctx -a claude-code -g -y

安装 repo 全部 skill

npx skills add AjayIrkal23/agentic-mercy-10x --all -g -y

预览 repo 内 skill

npx skills add AjayIrkal23/agentic-mercy-10x --list

SKILL.md

Frontmatter

{
    "name": "lean-ctx",
    "description": "Context Engineering for AI Agents — 81 MCP tools, 10 read modes, 95+ shell patterns, tree-sitter AST for 27 languages. Compresses LLM context by up to 99%. Use when reading files, running shell commands, searching code, or exploring directories. Auto-installs if not present."
}

lean-ctx — Context Engineering for AI Agents

Setup

which lean-ctx || curl -fsSL https://raw.githubusercontent.com/yvgude/lean-ctx/main/skills/lean-ctx/scripts/install.sh | bash
lean-ctx setup

Core Tools (10 always visible)

Tool	Purpose
`ctx_read(path, mode)`	Read file with compression and caching
`ctx_search(pattern, path)`	Search code with compressed results
`ctx_shell(command)`	Run shell with compressed output
`ctx_tree(path, depth)`	Directory listing
`ctx_patch(path, ops)`	Anchored editing (line+hash, no old-text echo)
`ctx_session(action)`	Session state and persistence
`ctx_knowledge(action)`	Project knowledge across sessions
`ctx_overview(task)`	Task-relevant project map
`ctx_graph(action)`	Code relationships and impact
`ctx_call(name, args)`	Invoke any tool by name

Shell Hook (use instead of raw exec)

lean-ctx -c "git status"
lean-ctx -c "cargo test"
lean-ctx -c "npm install"
lean-ctx ls src/

ctx_read Modes

Mode	When
`anchored`	Files you will edit (full text + `N:hh\|` anchors for ctx_patch)
`full`	Verbatim cached read
`map`	Context-only (deps + exports)
`signatures`	API surface only
`diff`	After edits (changed lines)
`aggressive`	Large files, syntax-stripped; JSON arrays row-deduped (lossless)
`entropy`	Shannon filtering
`task`	Task-relevant lines
`lines:N-M`	Specific range
`auto`	System selects optimal

Re-reads cost ~13 tokens. fresh=true bypasses cache. Redundant JSON (arrays of like objects) is crushed losslessly into a compact _defaults + per-row form; if a slice was dropped, recover it with ctx_expand(id, json_path=… | search=…).

File Editing

Anchored editing saves output tokens: ctx_read(mode="anchored") → ctx_patch(path, op, line, hash, new_text). Never reproduce old text byte-for-byte; batch via ops:[…]; op=create writes new files. Stale anchor → CONFLICT with fresh anchors (retry once). Native Edit/StrReplace stay fine; ctx_edit (str_replace) is the legacy fallback via ctx_call/power profile.

More Tools (via ctx_call or ctx_load_tools)

Architecture: ctx_symbol, ctx_callgraph, ctx_impact, ctx_architecture, ctx_routes, ctx_smells, ctx_quality ↳ "What breaks if I change this file/class/type?" → ctx_impact (file-level blast radius; resolves same-package/namespace type usage with no import for C#, Java, Go and Kotlin). "Who calls this function?" → ctx_callgraph (symbol-level). "How navigable / how much is complexity costing me?" → ctx_quality (navigability score + token quality tax). Multi-agent: ctx_agent, ctx_share, ctx_task, ctx_handoff, ctx_workflow Verify: ctx_benchmark, ctx_verify, ctx_proof, ctx_review Batch: ctx_fill, ctx_execute, ctx_expand, ctx_pack

Full docs: https://leanctx.com/docs