LLMCOM Prompt Compression

Compress prompts for Claude Code and Cursor CLI. Reduces tokens by removing filler words and abbreviating terms. Self-contained, no external dependencies req...

Security Scan

Instruction-only prompt-compression guide that matches its stated purpose and requests no installs, credentials, or special privileges.

Purpose & Capability

Name/description claim to compress prompts for Claude Code and Cursor CLI. SKILL.md contains only rules for removing filler words, abbreviations, and formatting examples consistent with that purpose; no unrelated capabilities are requested.

Instruction Scope

Runtime instructions are limited to text transformation rules and example formats. There are no commands to read files, access environment variables, call external endpoints, or collect system data.

Install Mechanism

No install spec and no code files — the skill is instruction-only, so nothing is downloaded or written to disk.

Credentials

The skill declares no environment variables, credentials, or config paths and the instructions do not reference any secrets; requested access is proportionate (none).

Persistence & Privilege

always is false and there is no indication the skill modifies agent config or demands permanent presence. The default ability for the agent to invoke the skill is normal and not problematic here.

Assessment

This is a low-risk, instruction-only skill that describes manual text-compression rules and requires no installs or secrets. Before using: (1) test outputs—aggressive removal of filler words (e.g., removing 'is', 'in', 'to') can change meaning or break LLM parsing; (2) verify the abbreviated tokens and pipe format are accepted by your target agent/CLI (Claude Code, Cursor) so you don't lose intent; (3) remember the skill is agent-invocable by default, so an agent could apply these rules automatically if allowed—ensure that automated rewriting is acceptable for your workflow. If you need automated, repeatable behavior consider implementing a tested code-based transformer rather than manual rule application.

Self-contained prompt compression tool. No installation required - just read this skill and apply the compression rules.

Compression Rules

1. Remove filler words: the, a, an, is, are, with, of, to, in, on, at, and, for, that, this

2. Abbreviate terms:

   • refactor -> refac
   • review -> rev
   • function -> func
   • security -> sec
   • performance -> perf
   • module -> mod
   • component -> comp
   • authentication -> auth
   • implementation -> impl

3. Output format for Claude Code: agent[task]|scope|focus|action|context|constraints

Example

Input: "Refactor the authentication module to improve type security"

Output: agent[refac]|mod|sec|review|@src|strict

Stats: 11 -> 6 tokens (45% reduction)

How to Use

Apply these rules manually when writing prompts for Claude Code CLI or Cursor agent.

1. Identify task type: refactor, review, fix, test, audit
2. Remove filler words
3. Abbreviate key terms
4. Format as pipe-separated values

Quick Reference

Task	Format
Refactor	agent[refac]
Security audit	agent[audit]
Bug fix	agent[fix]
Test coverage	agent[test]

Integration

Use before sending prompts to:

• Claude Code CLI (gent command)
• Cursor agent
• Any LLM API to reduce token costs