Bỏ qua tới nội dung
TakoAPI
Quay lại kỹ năng

compression

bởi trinitybotserver

AI & LLMs
—Chưa có đánh giá
Đánh giá của bạn

Tóm tắt

Pre-run prompt compression for iterative AI dev loops.

Mô tả

Kỹ năng này làm gì

A local script that compresses prompt files before AI dev loops run, reducing input token counts on each iteration. It preserves code blocks, URLs, file paths, and variable names. Creates .bak backups for instant undo.

Runs entirely locally with no model calls, so token savings happen at zero API cost.

Khi nào nên dùng

  • Cutting token costs before running long agent loops
  • Batch-compressing skill directories before iterative dev sessions
  • Reverting prompt changes when compressed output loses clarity
  • Adding prompt optimization as a Makefile step in AI-heavy repos
  • Trimming prompt files before submitting to token-billed API providers

Quy trình ví dụ

ĐẦU VÀO

User asks: compress all prompts in my agent directory before running the dev loop

AGENT
  1. 1

    Installs trinity-compress.config.json with target paths and compression rules

  2. 2

    Copies scripts/trinity-compress.sh into the repo

  3. 3

    Runs bash scripts/trinity-compress.sh balanced

  4. 4

    Reports token reduction and lists .bak backups created

ĐẦU RA

Prompt files compressed with .bak backups saved; ready to run the agent loop with reduced token usage

Cài đặt

clawhub install trinitybotserver/compression

Chạy lệnh CLI, hoặc dán prompt vào Claude Code, Cursor, Codex hay coding agent bạn đang dùng.

Liên kết