Skill compress
Token-efficient Agent Skills for Claude — Pillar 1: meta-skills ≤650t, Pillar 1.5: coding discipline
npx -y skills add Vix0007/vixero-skills --skill skill-compressAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Rewrite a verbose SKILL.md or prompt into its lean equivalent without changing behavior. Use when the user asks to compress, shrink, tighten, minify, rewrite-lean, or optimize the token count of a skill or prompt. Preserves triggers, output format, and hard rules. Reports before/after token counts.
SKILL.md
2.2 KB, as published. Nobody here has run it
skill-compress
Rewrite input at lowest token count that preserves behavior. Do not invent new behavior.
Trigger on
- "compress this skill / prompt"
- "make this leaner / tighter"
- "rewrite for fewer tokens"
- After
token-auditwhen user asks to apply the cuts
Rules, in order
- Frontmatter is sacred. Never change
name. Only improvedescriptionif it fails the description-craft checklist. Keep description 100–250 chars. - Strip preamble. Delete "I'll help", "Certainly", "Let me", "Sure", "Of course".
- Delete ceremony. "Please", "kindly", "feel free", "if you'd like".
- Imperative voice. "Do X" not "You should do X" or "It is recommended that X be done."
- Prose → table when a table saves 30%+ tokens for the same information.
- Lists over paragraphs for enumerations of three or more.
- Collapse examples. Keep one best example per concept. Move extras to an
examples/subfolder referenced by path. - Merge redundant sections. If "Guidelines" and "Rules" say similar things, merge.
- Code blocks stay. Never compress code. Compression errors in code cost more than tokens saved.
- Output format blocks stay verbatim. These are contracts.
Output
Return two sections:
1. Rewritten SKILL.md
Full content, ready to paste.
2. Diff summary
before: {n} tokens
after: {n} tokens
cut: {n} tokens ({pct}%)
preserved:
- trigger keywords
- output format
- hard rules
- behavior
changed:
- {specific change 1}
- {specific change 2}
Hard rules
- Never drop a hard rule. Rewrite it shorter.
- Never drop a trigger keyword from the description.
- If cut < 15% after compression, append: "Already lean. Structural rewrite needed for more savings."
- If user says "more aggressive" after first pass, apply
lean-responseregister as well: drop all softening, fragment sentences are fine.