caveman

A communication mode that shortens ordinary explanations while keeping code, numbers, paths, commands, and errors unchanged. It is enabled by default and pauses for safety warnings or questions needing a fuller explanation.

In plain words
What is it for?
Writing compressed progress updates, error summaries, and command results during coding work.
Why use it?
It reduces repeated wording and filler when brief status updates are enough.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/datit309/supergraph/caveman
Any agent
npx skills add datit309/supergraph --skill caveman
Clone the repo
git clone --depth 1 https://github.com/datit309/supergraph

Made for: Claude Code, Codex.

Per session 29 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 340 The whole file, excluding the scripts and references it only reads on demand.
Security scan C 1 finding. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00029 $0.00340
Opus 5 $0.00015 $0.00170
Sonnet 5 $0.00006 $0.00068
Haiku 4.5 $0.00003 $0.00034

Measured yesterday against content hash a137f86ed074, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade C, and why

caveman scanned grade C with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Recursive force deletehighDestructive command

rm -rf with a variable or a broad path is one typo away from removing the wrong tree.

- Multi-step destructive operations (git reset, rm -rf, etc.)
plugins/supergraph/skills/caveman/SKILL.md · 41 lines

What it actually says

/supergraph:caveman

Always-on compressed communication. Fewer words, same signal.

Announce: "🦴 caveman mode ON (always)"

Rules When Active

Strip: articles (a, an, the), filler phrases ("I'll now", "Let me", "Great question", "As you can see"), prepositions where unambiguous, subject pronouns ("I" at start of sentence).

Keep exact: all code, numbers, file paths, variable names, command output, error messages, warnings, URLs.

Structure: prefer bullet lists and tables over prose paragraphs. One clause per bullet. Skip transitional sentences.

Compress examples:

Normal Caveman
"I'll now run the tests to verify the fix works." "Running tests..."
"The error appears to be in the authentication module." "Error: auth module."
"I've updated the file and the tests are now passing." "Updated. Tests: PASS."

Auto-suspend (revert to normal temporarily)

Caveman mode suspends automatically for:

  • Safety warnings or security notices
  • Multi-step destructive operations (git reset, rm -rf, etc.)
  • Error explanations requiring precision
  • User asks a question requiring a full answer

Resume caveman after the suspended section automatically.

Mode

Always-on by default — no activation/deactivation needed. SUPERGRAPH_CAVEMAN flag is legacy, ignored.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 41 lines · 29 tokens per session scan C a137f86ed074

Subscribe to this mod's changes

caveman is a skill published in the GitHub repository datit309/supergraph (21 stars, last pushed 4d ago), licensed MIT. It adds 29 tokens to every session and 340 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

playwright-cli

Automates browser interactions for testing and validating your own web applications using playwright-cli. Use when you need terminal-first browser control for navigation, form filling, screenshots, tracing, bound browser sessions, debugging, or generating Playwright test code. Only use against applications you own or…

testdino-hq/playwright-skill · 64 tokens

migrate

Inventory an existing spec-driven project and apply bounded migrations to jig defaults: report, adopt-layout, rename-decisions, split-slices, slice-to-spec, seed-decisions, and copy-machinery. Use when the user says migrate this project to jig, adopt jig here, this repo already has specs — set up jig, scaffold-init…

ramboz/jig · 151 tokens

adr-workflow

Scaffold, accept, index, and link Architectural Decision Records (ADRs). Use when the user says "write an ADR", "record this decision", "resolve [deferred item] with an ADR", "supersede ADR-NNNN", or otherwise wants to capture a hard-to-reverse decision in docs/decisions/. Also use when a refinement-todo entry needs…

ramboz/jig · 122 tokens

bug-fix

Drive the teeth-gated lifecycle for reported defects: diagnose root cause, prove it, and prevent regression through REPORTED → DIAGNOSING → ROOTCAUSED → FIXING → REVIEWED → DONE, with VERIFIED, ESCALATED, and RESOLVEDONMAIN paths where needed. Auto-fires on fix this bug, debug this, root-cause this, this regressed…

ramboz/jig · 208 tokens

analyze

Cross-artifact consistency report for jig specs — a non-destructive six-category audit at CRITICAL/HIGH/MEDIUM/LOW severity, covering duplication, ambiguity, underspecification, principle violations, coverage gaps, and terminology drift. Auto-triggers when you say analyze this spec, check for inconsistencies, audit…

ramboz/jig · 160 tokens

arch-review

Team baseline for architecture, design-doc, and RFC review — produces summary, strengths, concerns, and open questions. Auto-triggers when you say review this design, review this architecture, review this proposal, review my RFC, poke holes in this proposal, is this design sound, or critique this tech spec. Defers to…

ramboz/jig · 207 tokens