ワンクリックでManusで任意のスキルを実行

prompt-compress

スター1

フォーク0

更新日2026年5月30日 21:37

Compress a system prompt or skill body while preserving semantic content. More aggressive than caveman:compress — targets system prompts and tool descriptions, validates with eval set. Triggers on "compress this prompt", "shrink system prompt", "reduce prompt tokens".

インストール

Codex または Claude でインストールこの Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。

Manusで実行

ソース

Andersonlimahw

Andersonlimahw/ai-starter-kit

GitHub リポジトリを開く Creator のリポジトリを見る

ダウンロード

Manusで実行

prompt-compress

Workflow

Read target prompt + eval set (REQUIRED — refuse without one).
Baseline: run eval, record score + token count.
Apply compressions iteratively (caveman style):
- Drop articles, fillers, pleasantries.
- Merge redundant rules.
- Convert prose lists → bullet lists.
- Inline examples if they replace explanation.
- Replace verbose phrasing with technical terms.
Re-run eval after each pass. Stop if score drops >2%.
Output diff + new prompt + score+token deltas.

Refuse to compress

Safety / refusal instructions (those stay verbose).
Examples (they're cheap calibration).
Anything user marks <keep>.

Output

Original:  4,820 tokens, eval 0.87
Pass 1:    3,100 tokens, eval 0.87  (-36%)
Pass 2:    2,400 tokens, eval 0.85  (-50%)  ← chose this
Pass 3:    1,900 tokens, eval 0.78  ✗ regression

Saved 50%. New eval within tolerance.

Companion

caveman:compress (memory files), eval-first (gate).

このリポジトリの他の Skills

同じリポジトリ

agent-loop-trace

Andersonlimahw/ai-starter-kit

Visualize an agent's perceive→think→act loop. Inspects tool calls, latency per step, decision branches, retry patterns. Surfaces bottlenecks. Triggers on "trace agent", "why is agent slow", "agent loop analysis", "what did the agent do".

2026-05-301

context-budget

Andersonlimahw/ai-starter-kit

Compute context window budget for a prompt/agent — sums system prompt, tool descriptions, examples, conversation history; flags overruns; suggests cuts. Triggers on "context budget", "how much context", "will this fit", "budget my prompt".

2026-05-301

data-is-code

Andersonlimahw/ai-starter-kit

Treat dataset / eval JSONL as code — lint, validate schema, detect duplicates, leakage between train/eval, label imbalance, PII. Karpathy "Software 2.0" lens. Triggers on "audit dataset", "lint jsonl", "check for leakage", "validate eval data".

2026-05-301

eval-first

Andersonlimahw/ai-starter-kit

Enforce "evals over vibes" — before any prompt, agent, or model change, require an eval set with success criteria and baseline measurement. Blocks vibe-only changes. Triggers on "tune prompt", "improve agent", "change model" without prior eval set.

2026-05-301

eval-set-builder

Andersonlimahw/ai-starter-kit

Build a 10-30 case eval dataset from a PRD, skill, or prompt — covering golden path, edge cases, and adversarial inputs. Output as JSONL with input/expected/category. Triggers on "build eval set", "create evals", "generate test cases for prompt".

2026-05-301

first-principles

Andersonlimahw/ai-starter-kit

Break down any ML/AI concept to first principles — strip jargon, derive from basics, build physical/mathematical analogy. Karpathy intuition-building style. Triggers on "first principles", "explain like I'm 5", "derive from scratch", "why does X actually work".

2026-05-301

name	prompt-compress
description	Compress a system prompt or skill body while preserving semantic content. More aggressive than caveman:compress — targets system prompts and tool descriptions, validates with eval set. Triggers on "compress this prompt", "shrink system prompt", "reduce prompt tokens".

prompt-compress

Workflow

Read target prompt + eval set (REQUIRED — refuse without one).
Baseline: run eval, record score + token count.
Apply compressions iteratively (caveman style):
- Drop articles, fillers, pleasantries.
- Merge redundant rules.
- Convert prose lists → bullet lists.
- Inline examples if they replace explanation.
- Replace verbose phrasing with technical terms.
Re-run eval after each pass. Stop if score drops >2%.
Output diff + new prompt + score+token deltas.

Refuse to compress

Safety / refusal instructions (those stay verbose).
Examples (they're cheap calibration).
Anything user marks <keep>.

Output

Original:  4,820 tokens, eval 0.87
Pass 1:    3,100 tokens, eval 0.87  (-36%)
Pass 2:    2,400 tokens, eval 0.85  (-50%)  ← chose this
Pass 3:    1,900 tokens, eval 0.78  ✗ regression

Saved 50%. New eval within tolerance.

Companion

caveman:compress (memory files), eval-first (gate).