How does it compress the guide?

It splits the guide into individual rules, strips filler and duplicate guidance, rewrites each as a short imperative, and keeps adding rules in priority order until your token budget is reached. The result is a tight list of actionable instructions rather than prose.

Style rules compete with your actual content for context window and cost money on every call. A budget forces the block down to the rules that change the output, so you are not paying to repeat a 2,000-word guide on every request.

Will it lose important rules?

It keeps rules in the priority order you set, so the most important sections survive truncation. If the budget is too tight for everything, raise it or tighten the source guide — the tool tells you how many rules it kept versus dropped.

Where should I put the block?

A system message is ideal because it persists across the conversation. If your tool has no system role, prepend the block to the user prompt above your actual instruction.

What is the Style Guide Prompt Injector?

Paste your brand or editorial style guide and get a compact, deduplicated style-instruction block you can prepend to any writing prompt. Pick a token budget and which sections matter most so the rules fit your context window. It runs free in your browser on Gera Tools, with nothing uploaded.

Style Guide Prompt Injector

Name: Style Guide Prompt Injector
Creator: Gera Tools
License: https://creativecommons.org/licenses/by/4.0/

Style guide prompt injector

Brand voice docs are written for humans — long, hedged, full of examples and rationale. An LLM does not need any of that; it needs a short list of concrete rules it can follow. Pasting the whole guide on every request burns context and money while burying the instructions that actually change the output. This tool turns a sprawling style guide into a tight, injectable block of imperatives that fits a token budget you choose.

How it works

The tool breaks your guide into individual rules — one per line or bullet — then cleans each one: it drops filler (“we generally try to”), removes near duplicates, and rewrites the rule as a short imperative (“Use active voice”). It then assembles the block in priority order, adding rules until your token budget is reached, so the sections you marked as most important survive even if everything cannot fit. You get a compact block plus a count of how many rules were kept and dropped, runnable as a system message or prompt prefix.

Why not just paste the whole style guide?

Three reasons. First, token cost: a 2,000-word style guide injected on every generation call adds roughly 500 tokens per request. At 50,000 calls a month and $2 per million input tokens, that is $50/month spent on rationale and examples the model does not need. Second, attention: models weight later instructions more heavily than earlier ones, so a long block of style guidance buries the most important rules under boilerplate. Third, context competition: every token your style block takes is a token unavailable for the actual content the model needs to process.

A 100-token compressed block of imperatives costs a tenth as much and often produces cleaner output because the rules are unambiguous.

What makes a style rule compressible

Some rules compress well; others do not. This affects how to structure the source guide:

Good for compression: short, categorical rules — “Write in second person”, “Avoid jargon”, “Sentences under 20 words”. These become one-line imperatives.
Hard to compress: rules that depend on examples — “Sound like our brand manifesto, see page 12 for the tone”. The tool cannot carry the example, only the instruction it implies.
Redundant: many guides restate the same idea five ways (“be concise”, “cut unnecessary words”, “avoid padding”, “keep it short”, “trim fluff”). The deduplication step collapses these to one rule, freeing budget for distinct guidance.

Before using this tool, a quick manual tidy of the source guide pays dividends: rewrite prose as bullets, remove worked examples, and consolidate repeated ideas.

Tips and example

Feed it bullets, not essays. A guide already written as short rules compresses cleanest. If yours is prose, the tool will still extract rules but you will get more from it after a quick tidy.
Prioritise voice over formatting. Tone and word-choice rules change output the most; spacing and punctuation conventions matter less to a model. Rank accordingly when the budget is tight.
Use it as a system message. Persisting the block in the system role keeps the style on across a long conversation without re-sending it each turn.
Tune the budget to the job. Short social copy can spare more budget for style than a long document where rules compete with your content.
Refresh it seasonally. Brand voice evolves. Re-run the injector when the source guide changes rather than manually editing the cached block.