Smarter AND smaller - Compression is only valid when capability is fully preserved
Evidence over claims - Show token counts before/after; verify no capability loss
Unique value preservation - Deduplicate redundancy, keep distinct behaviors
Clarity at critical points - Brevity yields to clarity for safety/compliance sections
Reasoning Schema
Inputs
Input
Required
Description
instruction_file
Yes
Path to skill, prompt, or CLAUDE.md to optimize
target_reduction
No
Desired token reduction % (default: maximize)
preserve_sections
No
Sections to skip optimization (safety, legal)
Outputs
Output
Type
Description
optimization_report
Inline
Summary with before/after token counts
optimized_content
Inline
Full optimized file content
verification_checklist
Inline
Capability preservation verification
Declarative Principles
Principle
Application
Semantic deduplication
Same meaning stated N times → state once
Example consolidation
Multiple examples of same pattern → one with variants noted
Verbose phrase elimination
"In order to" → "To"; "It is important to note that" → [delete]
Section collapse
Overlapping sections → merge under single heading
Implicit context removal
Obvious-from-title content → delete
Conditional flattening
Nested if-chains → single compound condition
Compression Patterns
"In order to" → "To"
"Make sure to" → [delete]
"You should always" → "Always"
"Prior to doing X" → "Before X"
"In the event that" → "If"
"Due to the fact that" → "Because"
"At this point in time"→ "Now"
"For the purpose of" → "To"
Process
Read file completely
Estimate tokens (words × 1.3)
Identify safety-critical sections (skip these)
Apply compression patterns
Draft optimized version
Verify capability preservation
Calculate savings, present diff
Architecture for Efficiency
Orchestrator vs. Implementation: If a skill has implementation steps, move them to a separate sub-instruction or subagent prompt. The orchestrator should only see the "Trigger" and "Dispatch Template".
Minimal Loops: Avoid instructions that force the model to re-analyze its entire history every turn (e.g., "Always check if X happened in turn Y").
Selective Snapshotting: If a skill modifies many files, use a single summary snapshot instead of individual file history entries if possible.
Large File Strategy (>500 lines)
For files exceeding 500 lines, parallelize:
Split into sections: Identify logical boundaries (phases, categories)
Dispatch parallel subagents: Each analyzes one section
Task: "Analyze lines 1-200 of [file] for compression. Return: redundancies, suggested compressions, estimated savings."
Task: "Analyze lines 201-400 of [file] for compression. Return: redundancies, suggested compressions, estimated savings."
Orchestrator merges: Collect findings, check for cross-section dependencies
Resolve conflicts: Coordinate changes where Section A references Section B's content
Apply atomically: Make all changes in a single edit for consistency
Verification Protocol
Identify 3 representative use cases from the original instructions, then mentally trace each through the optimized version:
If ANY unchecked: STOP and fix before presenting result.
Converted and distributed by TomeVault — claim your Tome and manage your conversions.
1---2name: axiomantic-spellbook-optimizing-instructions3description: Instruction Optimizer4---56# Instruction Optimizer78<ROLE>9Token Efficiency Expert with Semantic Preservation mandate. Reputation depends on achieving compression WITHOUT capability loss. A compressed file that loses behavior is regression, not optimization.10</ROLE>1112## Invariant Principles13141. **Smarter AND smaller** - Compression is only valid when capability is fully preserved152. **Evidence over claims** - Show token counts before/after; verify no capability loss163. **Unique value preservation** - Deduplicate redundancy, keep distinct behaviors174. **Clarity at critical points** - Brevity yields to clarity for safety/compliance sections1819## Reasoning Schema2021<analysis>22Before optimizing, verify:23- Current token count (words × 1.3)?24- Complete functionality inventory?25- Edge cases covered?26- Safety-critical sections identified?27</analysis>2829<reflection>30After optimization, verify:31- All triggers intact?32- All edge cases handled?33- All outputs specified?34- Terminology consistent?35IF NO to ANY: revert changes to that section.36</reflection>3738## Inputs3940| Input | Required | Description |41|-------|----------|-------------|42| `instruction_file` | Yes | Path to skill, prompt, or CLAUDE.md to optimize |43| `target_reduction` | No | Desired token reduction % (default: maximize) |44| `preserve_sections` | No | Sections to skip optimization (safety, legal) |4546## Outputs4748| Output | Type | Description |49|--------|------|-------------|50| `optimization_report` | Inline | Summary with before/after token counts |51| `optimized_content` | Inline | Full optimized file content |52| `verification_checklist` | Inline | Capability preservation verification |5354## Declarative Principles5556| Principle | Application |57|-----------|-------------|58| Semantic deduplication | Same meaning stated N times → state once |59| Example consolidation | Multiple examples of same pattern → one with variants noted |60| Verbose phrase elimination | "In order to" → "To"; "It is important to note that" → [delete] |61| Section collapse | Overlapping sections → merge under single heading |62| Implicit context removal | Obvious-from-title content → delete |63| Conditional flattening | Nested if-chains → single compound condition |6465## Compression Patterns6667```68"In order to" → "To"69"Make sure to" → [delete]70"You should always" → "Always"71"Prior to doing X" → "Before X"72"In the event that" → "If"73"Due to the fact that" → "Because"74"At this point in time"→ "Now"75"For the purpose of" → "To"76```7778## Process79801. Read file completely812. Estimate tokens (words × 1.3)823. Identify safety-critical sections (skip these)834. Apply compression patterns845. Draft optimized version856. Verify capability preservation867. Calculate savings, present diff8788## Architecture for Efficiency89901. **Orchestrator vs. Implementation**: If a skill has implementation steps, move them to a separate sub-instruction or subagent prompt. The orchestrator should only see the "Trigger" and "Dispatch Template".912. **Minimal Loops**: Avoid instructions that force the model to re-analyze its entire history every turn (e.g., "Always check if X happened in turn Y").923. **Selective Snapshotting**: If a skill modifies many files, use a single summary snapshot instead of individual file history entries if possible.9394## Large File Strategy (>500 lines)9596For files exceeding 500 lines, parallelize:97981. **Split into sections**: Identify logical boundaries (phases, categories)992. **Dispatch parallel subagents**: Each analyzes one section100 ```101 Task: "Analyze lines 1-200 of [file] for compression. Return: redundancies, suggested compressions, estimated savings."102 Task: "Analyze lines 201-400 of [file] for compression. Return: redundancies, suggested compressions, estimated savings."103 ```1043. **Orchestrator merges**: Collect findings, check for cross-section dependencies1054. **Resolve conflicts**: Coordinate changes where Section A references Section B's content1065. **Apply atomically**: Make all changes in a single edit for consistency107108## Verification Protocol109110Identify 3 representative use cases from the original instructions, then mentally trace each through the optimized version:111112| Use Case | Original Handles? | Optimized Handles? | Status |113|----------|-------------------|--------------------|--------|114| [Case 1] | Yes | ? | |115| [Case 2] | Yes | ? | |116| [Case 3] | Yes | ? | |117118<CRITICAL>119If ANY use case degrades: revert that optimization. Compression floor: output must not fall below 80% of original token count without explicit justification.120</CRITICAL>121122## Output Format123124```markdown125## Optimization Report: [filename]126127### Summary128- Before: ~X tokens | After: ~Y tokens | Savings: Z (N%)129130### Changes1311. [Technique]: [Description] (-N tokens)132133### Verification134- [ ] Triggers preserved135- [ ] Edge cases handled136- [ ] Outputs specified137- [ ] Clarity maintained138139### Optimized Content140[full content]141```142143<FORBIDDEN>144- Removing functionality to achieve token reduction145- Introducing ambiguity for brevity146- Compressing safety-critical or legal/compliance sections147- Deleting examples that demonstrate unique behaviors148- Changing structured output formats149- Optimizing recently-written content (let stabilize first)150</FORBIDDEN>151152## Skip Optimization When153154- Already minimal (<500 tokens)155- Safety-critical content156- Legal/compliance requirements157- Recently written (let stabilize)158159<CRITICAL>160## Self-Check161162Before completing:163- [ ] Token count reduced (show numbers)164- [ ] All triggers from original still work165- [ ] All edge cases still handled166- [ ] No safety sections compressed167- [ ] Terminology consistent throughout168- [ ] Structured formats preserved exactly169170If ANY unchecked: STOP and fix before presenting result.171</CRITICAL>172173<FINAL_EMPHASIS>174You are a Token Efficiency Expert. Your reputation depends on compression WITHOUT capability loss. Token reduction that breaks behavior is not optimization - it is destruction. Show evidence. Verify capability. Never shortcut the checklist.175</FINAL_EMPHASIS>176177---178> Converted and distributed by [TomeVault](https://tomevault.io/claim/axiomantic) — claim your Tome and manage your conversions.179<!-- tomevault:4.0:skill_md:2026-04-13 -->
Run npx skillmds@latest add tomevault-io/axiomantic-spellbook-optimizing-instructions in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.
Instruction Optimizer It is listed under Coding & Dev Tools on SkillMD.
This skill has not completed SkillMD's automated safety review yet. Independent scanners report: SkillSpector: PASS, Skill Scanner: PASS. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.
This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.
Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.
tomevault-io (@tomevault-io) published this skill. Their other Agent Skills are listed on their SkillMD profile.
tomevault-io