Install any skill in seconds. Free to start, no credit card required.
Get Started Free →alexgreensh/tldr
Compress a document, thread, transcript, or pasted text into a scannable TL;DR.
Passed 18 of 24 test cases with the skill, and 13 of 24 without it.
and 8 more in the full report
9 fixed9 passed either way2 still failing4 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 9 test cases the skill fixed. Same model, same prompt, with and without the skill.
Can you condense this product design meeting transcript into a quick briefing? Give me a top-level summary line, main key points, any action items assigned, and unresolved questions from the discussion. Here is the text: [pasted transcript]
| Checked | Without skill | With skill |
|---|---|---|
| The response contains exactly one question. | Missed | Passed |
| The response contains no introductory or explanatory text after the question. | Passed | Passed |
The instructions your model reads when the skill loads.
Compress the content the user pointed you at (a pasted block, a file, a link, a thread, a transcript, or the current selection) into a briefing they can absorb in seconds. This is a transform on someone else's content, not a style for your own answers, so summarize what's there, add nothing, invent nothing.
If they didn't say what to compress, ask which one thing, then stop. One question, nothing after it.
TL;DR: one line that carries the whole thing. A reader who reads only this line has the gist and the outcome. If the source has a decision, deadline, or ask, that goes here, not below.
Key points, three to seven, each its own line, bold the subject then a short clause. Keep every load-bearing number, name, date, threshold, and condition exactly as written, source wording over your paraphrase when it changes the meaning. Drop throat-clearing, repetition, and filler.
Action items only if the source actually contains them: who owns what, by when. Skip the heading entirely when there are none, never pad it.
Optional Open questions / unclear: one line each for anything the source leaves genuinely ambiguous. Flag it, don't resolve it by guessing.
test cases passed, vs 13 / 24 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.