Receber por email
← Voltar para o catálogo

vibecosystem

AI software team for Claude Code - 138 agents, 295 skills, 73 hooks. Self-learning, multi-agent swarm, autonomous skill evolution.

507estrelas
Atualizado há 4 meses

Ver no GitHub ↗ Licença: MIT

Como adicionar

/plugin marketplace add vibeeval/vibecosystem

O comando exato pode variar conforme o repositório. Confira o README no GitHub.

Para o autor da skill

Cole no README do seu repo

Mostra que sua skill está catalogada na Skillteca, gera backlink e tráfego rastreável.

[![Skillteca](https://www.skillteca.com.br/api/badge/vibecosystem-mpsjk4ra/svg)](https://www.skillteca.com.br/skills/vibecosystem-mpsjk4ra)
#ai

Skills relacionadas

Ver todas de Desenvolvimento →

Alerta por categoria

Receba novas skills de Desenvolvimento toda segunda

1 email curto, só com as skills novas de Desenvolvimento. 4 minutos de leitura, sem spam, cancela com 1 clique.

Você confirma o email no primeiro envio. Sem spam. Cancela com 1 clique.

Ler descrição completa

TLDR Deep Analysis

Full 5-layer analysis of a specific function. Use when debugging or deeply understanding code.

Trigger

  • /tldr-deep <function_name>
  • "analyze function X in detail"
  • "I need to deeply understand how Y works"
  • Debugging complex functions

Layers

| Layer | Purpose | Command | |-------|---------|---------| | L1: AST | Structure | tldr extract <file> | | L2: Call Graph | Navigation | tldr context <func> --depth 2 | | L3: CFG | Complexity | tldr cfg <file> <func> | | L4: DFG | Data flow | tldr dfg <file> <func> | | L5: Slice | Dependencies | tldr slice <file> <func> <line> |

Execution

Given a function name, run all layers:

# First find the file
tldr search "def <function_name>" .

# Then run each layer
tldr extract <found_file>              # L1: Full file structure
tldr context <function_name> --project . --depth 2  # L2: Call graph
tldr cfg <found_file> <function_name>  # L3: Control flow
tldr dfg <found_file> <function_name>  # L4: Data flow
tldr slice <found_file> <function_name> <target_line>  # L5: Slice

Output Format

## Deep Analysis: {function_name}

### L1: Structure (AST)
File: {file_path}
Signature: {signature}
Docstring: {docstring}

### L2: Call Graph
Calls: {list of functions this calls}
Called by: {list of functions that call this}

### L3: Control Flow (CFG)
Blocks: {N}
Cyclomatic Complexity: {M}
[Hot if M > 10]
Branches:
  - if: line X
  - for: line Y
  - ...

### L4: Data Flow (DFG)
Variables defined:
  - {var1} @ line X
  - {var2} @ line Y
Variables used:
  - {var1} @ lines [A, B, C]
  - {var2} @ lines [D, E]

### L5: Program Slice (affecting line {target})
Lines in slice: {N}
Key dependencies:
  - line X → line Y (data)
  - line A → line B (control)

---
Total: ~{tokens} tokens (95% savings vs raw file)

When to Use

  1. Debugging - Need to understand all paths through a function
  2. Refactoring - Need to know what depends on what
  3. Code review - Analyzing complex functions
  4. Performance - Finding hot spots (high cyclomatic complexity)

Programmatic API

from tldr.api import (
    extract_file,
    get_relevant_context,
    get_cfg_context,
    get_dfg_context,
    get_slice
)

# All layers for one function
file_info = extract_file("src/processor.py")
context = get_relevant_context("src/", "process_data", depth=2)
cfg = get_cfg_context("src/processor.py", "process_data")
dfg = get_dfg_context("src/processor.py", "process_data")
slice_lines = get_slice("src/processor.py", "process_data", target_line=42)
Compartilhar X LinkedIn