L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Productivité · SKILL.md

tldr-overview

tldr-overview

Get a token-efficient overview of any project using the TLDR stack

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: tldr-overview description: Get a token-efficient overview of any project using the TLDR stack

TLDR Project Overview

Get a token-efficient overview of any project using the TLDR stack.

Trigger

  • /overview or /tldr-overview
  • "give me an overview of this project"
  • "what's in this codebase"
  • Starting work on an unfamiliar project

Execution

1. File Tree (Navigation Map)

tldr tree . --ext .py    # or .ts, .go, .rs

2. Code Structure (What Exists)

tldr structure src/ --lang python --max 50

Returns: functions, classes, imports per file

3. Call Graph Entry Points (Architecture)

tldr calls src/

Returns: cross-file relationships, main entry points

4. Key Function Complexity (Hot Spots)

For each entry point found:

tldr cfg src/main.py main  # Get complexity

Output Format

## Project Overview: {project_name}

### Structure
{tree output - files and directories}

### Key Components
{structure output - functions, classes per file}

### Architecture (Call Graph)
{calls output - how components connect}

### Complexity Hot Spots
{cfg output - functions with high cyclomatic complexity}

---
Token cost: ~{N} tokens (vs ~{M} raw = {savings}% savings)

When NOT to Use

  • Already familiar with the project
  • Working on a specific file (use targeted tldr commands instead)
  • Test files (need full context)

Programmatic Usage

from tldr.api import get_file_tree, get_code_structure, build_project_call_graph

# 1. Tree
tree = get_file_tree("src/", extensions={".py"})

# 2. Structure
structure = get_code_structure("src/", language="python", max_results=50)

# 3. Call graph
calls = build_project_call_graph("src/", language="python")

# 4. Complexity for hot functions
for edge in calls.edges[:10]:
    cfg = get_cfg_context("src/" + edge[0], edge[1])
Skills enregistrées