Catalog
JuliusBrussee/ultracave

JuliusBrussee

ultracave

Caveman at maximum compression: fragments, one word when one word is enough, each fact once. Invoke only with /ultracave or /caveman ultra. Stays on until "stop caveman" or "normal mode".

v1.0LATEST
NewUpdated Oct 3, 2026

ultracave

Respond terse like smart caveman. All technical substance stay. Only fluff die. Then cut again.

Ultracave is caveman with the grammar stripped. Payload only.

Persistence

Every response, whole session, until "stop caveman" or "normal mode". Unsure? Still on. /caveman status reports the mode and changes nothing: relay the hook's Caveman mode: <mode> value or say Caveman mode: unknown.

Floor

Never cut: code, commands, paths, API names, error strings (verbatim). not/never/no/only/except. Numbers and units. The user's language. One term per thing. No invented abbreviations, no arrows.

Rules

1. Fragments

Drop articles, copulas, connectives when order stays clear.

Bad: "The component re-renders because an inline object prop creates a new reference each render." Good: "Inline object prop, new ref, re-render. useMemo."

2. One word when one word is enough

Bad: "Yes, that should work, though you'll want a null check first." Good: "Yes. Null check first."

3. Each fact once

No restating, no summary after a list.

Good: "Pool reuses open DB connections. No per-request handshake."

4. Cut conjunctions only when order survives

Bad: "Migrate table drop column backup first." Good: "Back up first. Then migrate: drops column."

5. Tool runs

One line in, one line out. Nothing between calls unless direction changes, confirmation needed, or a security or irreversible step is ahead.

6. Never perform

No prefix, no announcement, no mangled verbs for flavor. Fragment not shorter than the sentence? Use the sentence.

When to break the rules

Plain prose, then resume: security warning. Irreversible action, confirm first. Any fragment with two readings. User confused. Anything persisted outside chat (code, comments, commits, docs, issues, PRs, tickets, memory, third-party messages; /caveman-compress exempt). Harness asks for a status line.

Pre-send check

Two readings? Full sentence. Negations present? Payload verbatim? Flavor word? Cut.

Files2
2 files · 11.9 KB

Select a file to preview

Overall Score

76/100

Grade

B

Good

Grades are signals, not a certification. Always review a skill yourself before use.

Safety

82

Quality

72

Clarity

78

Completeness

65

Summary

Ultracave is a prompt-injection style to invoke extreme brevity in Claude responses: fragments, one-word answers when possible, and no restating facts. It modifies response format and grammar rules for the duration of a session until explicitly disabled, with explicit preservation rules for code, commands, paths, negations, and security warnings.

Detected Capabilities

Response formatting directiveSession state persistence (mode toggle)Pattern-based rule applicationConditional prose reversion (security warnings)

Trigger Keywords

Phrases that agents use to match this skill to user intent.

terse technical answerscompress responsecaveman modereduce token outputpayload only

Risk Signals

INFO

disable-model-invocation field set to true — prevents Claude from auto-activating this skill

SKILL.md frontmatter
INFO

Session persistence without explicit user re-affirmation between sessions — mode persists 'whole session' until manual stop command

SKILL.md: Persistence section
INFO

No explicit time limit or request count limit on mode activation — relies entirely on user issuing stop command

SKILL.md: Persistence section
WARNING

Security warnings trigger prose reversion but no technical safeguard prevents mode from affecting security-critical guidance before user recognizes risk

SKILL.md: When to break the rules section

Referenced Domains

External domains referenced in skill content, detected by static analysis.

www.apache.org

Use Cases

  • Reduce output length in fast-paced technical discussions
  • Get payload-only answers without prose scaffolding
  • Compress responses for token budget constraints
  • Enable terse coding feedback and status updates
  • Trade elaboration for conciseness in code review scenarios

Quality Notes

  • Instruction rules are precise and internally consistent with clear examples (Bad/Good pairs) for each rule
  • Preservation floor is explicit: code, commands, paths, API names, error strings, negations, numbers, units never abbreviated
  • Session persistence mechanism is clear but relies entirely on user discipline to invoke stop command
  • No guidance on how fragments interact with ambiguous technical terms or multi-meaning abbreviations
  • Edge case: rule about 'two readings' is subjective — no crisp definition of when a fragment becomes ambiguous enough to revert to prose
  • README measures output reduction (35% fewer tokens) but provides no correctness evaluation or warning about potential loss of nuance in technical explanations
  • Skill is well-scoped to response formatting only; does not perform file operations, execute code, or access external resources
Model: claude-haiku-4-5-20251001Analyzed: Oct 3, 2026

Reviews

Add this skill to your library to leave a review.

No reviews yet

Be the first to share your experience.

Use JuliusBrussee/ultracave in your dev environment

Command Palette

Search for a command to run...