Create advanced Works with: Any AI

Prompt Compression Before and After

Before after Has example

Shorten a long prompt without losing behavior.

The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.

Compress my prompt while preserving results. The prompt: [PASTE_PROMPT]. Why I want it shorter (cost, speed, context limits, readability): [REASON]. The test cases that define good behavior: [TEST_CASES].

Return:
BEFORE: the prompt with annotations: [redundant], [vague filler], [duplicate rule], [example that does not teach anything new], [instruction that conflicts], [necessary]. Count words and note the sections.
AFTER: a compressed prompt that keeps necessary rules, merges duplicates, replaces long explanations with precise wording, trims examples to the most informative ones, and uses a consistent structure.
CHANGE LOG: each removed or merged part with the reason and the risk.
TEST PLAN: run both versions on the test cases and compare using a rubric; identify behaviors that disappeared; restore any rule whose removal caused regressions.
TIERS: a minimal version (essential rules only), a standard version, and a detailed version for difficult cases.

Warn about over-compression: ambiguity increases when instructions are cut too far. State that I must verify results because behavior is sensitive to wording. Do not change the safety or privacy rules.

Fill in before use

Optional: type your details and the prompt above updates. Nothing is saved or sent.

Replace each bracketed placeholder with your own details, then paste into Any AI or another assistant.

BEFORE (212 words): annotated - 3 duplicate rules on tone [duplicate rule], long background paragraph [redundant], 4 examples teaching the same pattern [example adds nothing new], one conflicting length rule.
AFTER (118 words): merged tone rules into one; kept two examples (one typical, one edge); replaced the background paragraph with a one-line context; one length rule.
Change log: removed example 3 and 4 (risk: edge case missing, covered by example 2).
Test: both versions on 12 cases; restore any rule whose removal caused regressions. Tiers: minimal, standard, detailed.

• compression