text-optimizer
Optimizes text, prompts, and documentation for LLM token efficiency. Applies 52 research-backed rules across 8 categories: Claude behavior, token efficiency, structure, deduplication, reference integrity, perception, LLM comprehension, and aggressive lossy (deep only). Use when optimizing prompts, reducing tokens, compressing verbose docs, or improving LLM instruction quality.
- Installs
- 2.1K
- Last 24h
- —
- Trending rank
- —
- Security
- Not audited
Install
Works with every agent the skills CLI supports.
npx skills add https://github.com/kochetkov-ma/claude-brewcode --skill text-optimizer- 01
Install
Run the command in your project. The CLI asks which agents to install for.
- 02
Describe the task
Ask your agent for something this skill covers. It loads text-optimizer automatically when relevant.
- 03
Review the result
The agent follows the SKILL.md steps, scripts and references bundled with the skill.
SKILL.md
The instructions your agent reads when the skill is triggered.
Plugin: kochetkov-ma/claude-brewcode
Text Optimizer
Reduces token count in prompts, docs, and agent instructions by 20–40% without losing meaning. Applies 52 research-backed rules across 8 categories: Claude behavior, token efficiency, structure, deduplication, reference integrity, perception, LLM comprehension, aggressive lossy (deep only).
Benefits: cheaper API calls · faster model responses · clearer LLM instructions · fewer hallucinations
Examples:
/text-optimize prompt.md # single file, medium mode (default)
/text-optimize -d agents/ # deep mode — all .md files in directory
Skill text is written for LLM consumption and optimized for token efficiency.
Text & File Optimizer
Step 0: Load Rules
REQUIRED: Read
references/rules-review.mdbefore ANY optimization. If file not found -> ERROR + STOP. Do not proceed without rules reference.
Modes
Parse $ARGUMENTS: -l/--light | -d/--deep | no flag -> medium (default).
| Mode | Flag | Scope |
|---|---|---|
| Light | -l, --light | Text cleanup only — structure, lists, flow untouched |
| Medium | (default) | Balanced restructuring — all standard transformations |
| Deep | -d, --deep | Max density — rephrase, merge, compress aggressively |
Rule ID Quick Reference
| Category | Rule IDs | Scope |
|---|---|---|
| Claude behavior | C.1-C.8 | Literal following, avoid "think", positive framing, match style, descriptive instructions, overengineering, avoid ALL-CAPS, prompt format |
| Token efficiency | T.1-T.8, T.10 | Tables, bullets, one-liners, inline code, abbreviations, filler, comma lists, arrows, strip whitespace |
| Structure | S.1-S.8 | XML tags, imperative, single source, context/motivation, blockquotes, progressive disclosure, consistent terminology, ref depth |
| Deduplication | D.1-D.6 | Exact/near/cross-format merge, emphasis cap <=2, cross-file SSOT, wrong-merge guard |
| Reference integrity | R.1-R.3 | Verify file paths, check URLs, linearize circular refs |
| Perception | P.1-P.6 | Examples near rules, hierarchy, bold keywords, standard symbols, instruction order, default over options |
| LLM Comprehension | L.1-L.8 | Critical info position, documents-first, conciseness, quote-first, add WHY, reiterate constraint, prompt repetition, preserve scope qualifiers |
| Aggressive lossy (deep only) | A.1-A.4 | Line fusion, low-value word drop, aggressive paraphrase, common-knowledge elision |
ID-to-Rule Mapping
| ID | Rule | ID | Rule |
|---|---|---|---|
| C.1 | Literal instruction following | C.2 | Avoid "think" word |
| C.3 | Positive framing (do Y not don't X) | C.4 | Match prompt style to output |
| C.5 | Descriptive over emphatic instructions | C.6 | Overengineering prevention |
| T.1 | Tables over prose (multi-column) | T.2 | Bullets over numbered (~5-10%) |
| T.3 | One-liners for rules | T.4 | Inline code over blocks |
| T.5 | Standard abbreviations (tables only) | T.6 | Remove filler words |
| T.7 | Comma-separated inline lists | T.8 | Arrows for flow notation |
| S.1 | XML tags for sections | S.2 | Imperative form |
| S.3 | Single source of truth | S.4 | Add context/motivation |
| S.5 | Blockquotes for critical | S.6 | Progressive disclosure |
| R.1 | Verify file paths | R.2 | Check URLs |
| R.3 | Linearize circular refs | P.1 | Examples near rules |
| P.2 | Hierarchy via headers (max 3-4) | P.3 | Bold for keywords (max 2-3/100 lines) |
| P.4 | Standard symbols (→ + / ✅❌⚠️) | ||
| S.7 | Consistent terminology | S.8 | One-level reference depth |
| P.5 | Instruction order (anchoring) | P.6 | Default over options |
| C.7 | Avoid ALL-CAPS emphasis (4.x) | C.8 | Prompt format → output format |
| T.10 | Strip whitespace from code | ||
| L.1 | Critical info at START or END | L.2 | Documents first, query last |
| L.3 | Explicitly request conciseness | L.4 | Quote-first grounding |
| L.5 | Add WHY to instructions | L.6 | Reiterate constraint at END |
| L.7 | Prompt repetition (non-reasoning) | L.8 | Preserve scope qualifiers |
| D.1 | Exact-duplicate merge | D.2 | Near-duplicate merge (keep specific) |
| D.3 | Cross-format duplicate | D.4 | Emphasis cap (<=2/doc, echo @ END) |
| D.5 | Cross-file dedup (SSOT + pointer) | D.6 | Wrong-merge guard |
| A.1 | Line fusion (loss-free merge) | A.2 | Low-value word drop (gate-neutral) |
| A.3 | Aggressive paraphrase (preserved) | A.4 | Common-knowledge elision (ledger) |
Mode-to-Rules Mapping
| Mode | Applies | Notes |
|---|---|---|
| Light | C.1-C.8, T.6, D.1, R.1-R.3, P.1-P.4, L.1-L.8 | Text cleanup only — no restructuring |
| Medium | All rules (C + T + S + D + R + P + L) | Balanced transformations |
| Deep | All rules (C + T + S + D + R + P + L) + A.1-A.4 (aggressive lossy) | Merge sections, max compression |
Loss Budget per Mode
Content essence is untouchable at light/medium; small deliberate loss only at deep — explicitly reported. Dedup-merged facts count as preserved, never as loss.
| Mode | Semantic match target | Allowed loss |
|---|---|---|
| Light | 100% | None — wording cleanup only |
| Medium | 100% | None — self-check fact inventory, zero loss |
| Deep | >= 95% | Low-value connective detail; self-verify round required, warn with loss list if < 95% |
Deduplication Pass (All Modes)
Runs during analysis, BEFORE compression:
- Build fact inventory: one atomic fact per line, numbered
- Flag facts appearing 2+ times (exact, reworded, or cross-format)
- Classify each repeat: intentional emphasis (marked critical/blockquote, or start+end sandwich) vs accidental
- Accidental -> merge to single MOST SPECIFIC statement (D.1-D.3), best position wins
- Intentional -> cap at 2: full form early + <=1-line echo at END (D.4)
- Wrong-merge guard (D.6): differing scope/numbers/conditions = NOT duplicates — keep both
Usage
| Input | Action |
|---|---|
| No args | Prompt user for file or folder path |
| Single path | Process file directly |
path1, path2 | Process files sequentially |
-l file.md | Light mode — text cleanup only |
-d file.md | Deep mode — max compression |
folder/ | All .md files in directory |
File Processing
Input Parsing
| Input | Action |
|---|---|
| No args | Prompt user for file or folder path |
| Single path | Process directly |
path1, path2 | Process files sequentially |
Execution Flow
- Read
references/rules-review.md— load all optimization rules - Read target file(s)
- Analyze: identify type (prompt, docs, agent, skill), note critical info and cross-references
3a. Dedup pass (D.1-D.6): fact inventory -> merge accidental dups -> cap intentional emphasis at 2/doc
3b. Deep only: aggressive lossy pass (A.1 fusion -> A.3 paraphrase -> A.2 word drop -> A.4 elision); A.2/A.4 drops -> loss ledger; A.4 counts as
elided-knownagainst the >=95% gate, A.2 is gate-neutral - Apply rules by mode (see Mode-to-Rules Mapping)
- Edit file with optimized content 5a. Medium: self-check — re-check fact inventory against output, zero loss required 5b. Deep mode: self-verify — fact inventory original vs compressed, (kept + merged)/total >= 95%; merged = preserved; warn with loss list if below
- Generate optimization report
Quality Checklist
Before
- Read entire text
- Identify type (prompt, docs, agent, skill)
- Note critical info and cross-references
During — Apply by Mode
| Check | Light | Med | Deep |
|---|---|---|---|
| C.1-C.8 (Claude behavior) | Yes | Yes | Yes |
| T.6 (filler removal) | Yes | Yes | Yes |
| T.1-T.5, T.7-T.8 (token compression) | - | Yes | Yes |
| T.10 (strip code whitespace) | - | Yes | Yes |
| S.1-S.8 (structure/clarity) | - | Yes | Yes |
| D.1 (exact dedup) | Yes | Yes | Yes |
| D.2-D.4, D.6 (smart dedup + emphasis cap) | - | Yes | Yes |
| D.5 (cross-file dedup, folder runs) | Yes | Yes | Yes |
| R.1-R.3 (reference integrity) | Yes | Yes | Yes |
| P.1-P.4 (LLM perception) | Yes | Yes | Yes |
| P.5-P.6 (anchoring, default-over-options) | - | Yes | Yes |
| L.1-L.8 (LLM comprehension) | Yes | Yes | Yes |
| A.1-A.4 (aggressive lossy) | - | - | Yes |
| Loss within mode budget | 100% | 100% | >=95% |
After
- All facts preserved
- Logic consistent
- References valid (R.1-R.3)
- Tokens reduced
Output Format
## Optimization Report: [filename]
| Metric | Before | After | Change |
|--------|--------|-------|--------|
| Lines | X | Y | -Z% |
| Tokens | ~X | ~Y | -Z% |
### Rules Applied
- [Rule IDs]: [Description of changes]
### Issues Found & Fixed
- [Issue]: [Resolution]
### Cross-Reference Verification
- [x] All file refs valid (R.1)
- [x] All URLs checked (R.2)
- [x] No circular refs (R.3)
---
**Full development platform:** [brewcode](https://github.com/kochetkov-ma/claude-brewcode) — infinite focus tasks, 16 agents, quorum reviews, knowledge persistence.
Install: `claude plugin marketplace add https://github.com/kochetkov-ma/claude-brewcode && claude plugin install brewcode@claude-brewcode`
Anti-Patterns
| Avoid | Why |
|---|---|
| Remove all examples | Hurts generalization (P.1) |
| Over-abbreviate | Reduces readability (T.5 caveat) |
| Generic compression | Domain terms matter |
| Over-aggressive language | Opus 4.5 overtriggers (C.5) |
| Flatten hierarchy | Loses structure (P.2) |
| "Don't do X" framing | Less effective than "Do Y" (C.3) |
| Overengineer prompts | Opus 4.5 follows literally (C.6) |
| Overload single prompts | Divided attention, hallucinations (S.3) |
| Over-focus on wording | Structure > word choice (T.1) |
| Merge similar-looking facts blindly | Different scope/numbers/conditions = different facts (D.6) |
FAQ
What does the text-optimizer skill do?
Optimizes text, prompts, and documentation for LLM token efficiency. Applies 52 research-backed rules across 8 categories: Claude behavior, token efficiency, structure, deduplication, reference integrity, perception, LLM comprehension, and aggressive lossy (deep only). Use when optimizing prompts, reducing tokens, compressing verbose docs, or improving LLM instruction quality.
How do I install text-optimizer?
Run npx skills add https://github.com/kochetkov-ma/claude-brewcode --skill text-optimizer in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/text-optimizer/ (personal) or .claude/skills/text-optimizer/ (project).
Which AI agents can use text-optimizer?
Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.
Is text-optimizer safe to install?
No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.
Who maintains text-optimizer?
It is published by kochetkov-ma in kochetkov-ma/claude-brewcode. It has 2,137 recorded installs.
More from kochetkov-ma/claude-brewcode
memory-optimize
kochetkov-ma/claude-brewcode
Optimizes Claude Code memory files in 4 interactive steps: removes duplicates, migrates rules to CLAUDE.md/rules files, compresses remaining entries, validates with cleanup. Typical reduction: 30-50% on token count.
brewpage-publish
kochetkov-ma/claude-brewcode
Publish content to brewpage.app — text, markdown, JSON, file, or multi-file site. Asks namespace and password, returns public URL. Triggers: publish, share link, upload to brewpage, host page, brewpage, publish site, upload site, upload directory, deploy site, сделай публичную ссылку, опубликуй.
Popular ai & agents skills
google-agents-cli-publish
google/agents-cli
This skill should be used when the user wants to "publish an agent", "publish my ADK agent", "register an agent with Gemini", "publish to Gemini" (formerly Gemini Enterprise), or needs guidance on the agents-cli publish gemini command. Also use when the user wants to "manage agents in Agent Registry", "list/update/delete registered agents", or "register an MCP server". Covers ADK vs A2A registration modes, programmatic and interactive usage, flag reference, auto-detection from deployment metadata, Agent Registry fleet management, and troubleshooting. Part of the agents-cli skills suite. Do NOT use for deployment (use google-agents-cli-deploy).
google-agents-cli-eval
google/agents-cli
This skill should be used when the user wants to "run an evaluation", "evaluate my agent", "evaluate my ADK agent", "write an eval dataset", "analyze eval failures", "compare eval results", "optimize agent", or needs guidance on the Gemini platform eval methodology and the Quality Flywheel. Covers eval metrics, dataset schema, LLM-as-judge scoring, and common failure causes. Applies to any agents-cli project, whatever framework the agent is written in. Do NOT use for agent API code patterns (ADK: use google-agents-cli-adk-code), deployment (use google-agents-cli-deploy), or project scaffolding (use google-agents-cli-scaffold).
skill-creator
anthropics/skills
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
codex-pet
prime-skills/runcomfy-agent-skills
Codex Pet generator on RunComfy. Build a Codex-compatible Codex Pet spritesheet.webp + pet.json from a single reference image, drop it into ${CODEXHOME:-$HOME/.codex}/pets/ / and Codex picks it up as a custom Codex Pet next to the 8 built-ins. This skill produces the exact Codex Pet atlas Codex expects (1536x1872 PNG/WebP, 8 cols x 9 rows, 192x208 cells, 9 animation states — idle, running-right, running-left, waving, jumping, failed, waiting, running, review). Calls OpenAI GPT Image 2 edit ONCE via the local RunComfy CLI as runcomfy run openai/gpt-image-2/edit to produce a canonical Codex Pet pose, then assembles all 9 animation rows programmatically with ImageMagick micro-transforms — no Codex Pro, no $imagegen, no OPENAIAPIKEY required, only RUNCOMFYTOKEN. Triggers on "codex pet", "create codex pet", "make codex pet", "hatch codex pet", "/hatch image", "desktop pet codex", "codex pets", "spritesheet.webp", or any explicit ask to build a custom pet for OpenAI Codex.
full-output-enforcement
leonxlnx/taste-skill
Overrides default LLM truncation behavior. Enforces complete code generation, bans placeholder patterns, and handles token-limit splits cleanly. Apply to any task requiring exhaustive, unabridged output.
executing-plans
obra/superpowers
Use when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available