AgentSkillsCN

Tldr First

TLDR 优先的工作流——在读取文件前先进行索引，以最大限度地减少 Token 使用。

SKILL.md
--- frontmatter
description: TLDR-first workflow - minimize tokens by indexing before reading files
globs: "**/*"
alwaysApply: true

TLDR-First Workflow

Always run TLDR commands before reading code files.

Quick Start

bash
tldr warm .                    # Index repo (run once)
tldr structure .               # Project structure
tldr arch .                    # Entry/leaf functions
tldr extract <file>            # Function signatures + call graph
tldr impact <function>         # Find all callers (reverse call graph)
tldr calls .                   # Project-level call graph
tldr cfg <file> <function>     # Control flow graph
tldr dfg <file> <function>     # Data flow graph

Workflow Rules

  1. •Never read full files — use TLDR extract first
  2. •Only open targeted ranges — sed -n 'start,endp' file when needed
  3. •Paste TLDR outputs — helps future agents understand context
  4. •Search before read — tldr extract shows line numbers

Example Flow

bash
# Task: Fix bug in upload handler
tldr extract src/main.rs | grep -A5 upload_file   # Find signature + calls
sed -n '172,240p' src/main.rs                      # Read only that function

# Task: Find escapeHtml usage in frontend
rg 'escapeHtml|escapeAttr' static/index.html      # Discover usage patterns
sed -n '100,120p' static/index.html               # Read targeted range

Frontend Note

WARNING: Never run tldr extract static/index.html — output exceeds 6000 lines and will blow up token budget. Use rg/grep to find specific patterns, then sed -n for targeted ranges.

Quality Checks (Pre-Commit)

bash
cargo build                # Must pass
cargo clippy               # No warnings
npm run test:e2e           # If UI changed

Security Reminders

  • •Backend: clean_relative_path() + resolve_path() for all user paths
  • •Frontend: escapeHtml()/escapeAttr() for all user content
  • •Broadcast: All mutations use broadcast_update()