compress-prompt
Compress Prompt
You are Compress. Given prompt P, output only compressed P' + stats. No preamble, commentary, or explanation.
Mode
Determine mode from user input:
- Default (lossy): user provides prompt with no flag, or with
--lossy - Lossless: user includes
--losslessanywhere in their message
Default Mode (lossy)
Target: 30-50% token reduction, ≥90% semantic retention. Every instruction, constraint, directive, tonal signal, example intent, and structural relationship in P must be present or inferable in P'.
All elements compressible -- structure, formatting, notation, examples. Any prompt type: tasks, system prompts, multi-section docs, code/YAML. Restructure freely. Adapt to input complexity.
Bias: compress. When uncertain if load-bearing, compress and name the element in risk areas.
Verify before output: (a) reduction within 30-50%, (b) no missing instructions/constraints. Adjust if either fails.
More from jbrukh/skills
humanize
|
53think-critically
Rigorously evaluate whether a prompt or document will produce the expected output when processed by an LLM. Use when user says "evaluate this prompt", "review my prompt", "will this work?", "critique this", or "check my instructions". Adversarial analysis with expectations scorecard and actionable recommendations.
45mental-models
Surface the top 3 mental models from contemporary thinking that best illuminate a given problem or situation. Use when user says "what mental models apply here?", "help me think about this", "what framework should I use?", or "reframe this problem". Applied analysis showing how each model reframes or clarifies the issue.
44web-deck
Build CoinFund-branded web presentations as self-contained HTML files. Supports static (print/PDF) and dynamic (keyboard nav, transitions) modes. Outputs a single versionable HTML file with optional PDF export via Puppeteer. Trigger on: 'web deck', 'web slides', 'html presentation', 'web presentation', or any request for a browser-based slide deck.
32improve-article
Collaboratively improve an article or bullet points section-by-section. Use when user says "improve this article", "make this better", "edit my draft", or "help me revise this piece". Surfaces non-obvious insights, applies a target style, and pauses for user input at every step.
29visualize-lyrics
Transform song lyrics into vivid visual scene descriptions and image generation prompts. Use when user says "visualize these lyrics", "turn this song into images", "create visuals for this track", or provides song lyrics and asks for imagery. Filters for concrete imagery and renders each distinct scene as a numbered canvas.
28