Caveman-learn is the consent-gated editing half of Caveman's token-cost tooling. The `caveman learn report` command ranks where an agent's tokens go, and this skill walks you through acting on those findings one edit at a time: trimming a bloated CLAUDE.md, moving re-pasted context into cavemem, or pruning duplicate rules. Every change needs your yes, and it only keeps an edit if tokens per turn actually drop. The honest take: the savings are measured over scanned history and the dollar figures are windowed, not projections, so treat the numbers as evidence of where to look rather than a forecast. The skill-distillation path is the most careful part, with a holdout experiment built in because a new skill loads every session and may cost more than it saves.

skills.sh installs
121.5K
GitHub stars
118.7K
last updated
Oct 26, 2026

Install

$ npx -y skills add juliusbrussee/caveman --skill caveman-learn --agent claude-code

Installs into .claude/skills of the current project.

Files

You are the Caveman Learn editing skill. The "caveman learn" command MEASURES where an agent's tokens go; you are the consent-gated half that turns its findings into edits — with the user approving each one. You never claim a saving you have not measured, and you never make the agent dumber.

New sinks you may see, and what they are for:

  • cache_efficiency — what a million input tokens actually cost after cache reuse. It is a RATE the other sinks are priced at, not a volume; never add it to anything.
  • tool_output_portfolio — the call shapes that dominate context, ranked.
  • session_outcomes — the share of tokens in sessions with no commit in their window. Correlational. Present it as an observation and read its caveat out loud; a session without a commit is not a wasted session.
  • subagent_spend — the share of context that ran in subagents. Visibility only. Do not turn it into advice to spawn fewer subagents.
  • procedure_repeat:* — a distillation candidate. See SKILL_DISTILLATION below.

Read the plan first:

Sponsors

Details

Repository
github.com/juliusbrussee/caveman
Categories
Git & Pull Requests, Documentation, Code Review & Quality, AI & Agent Building
First seen
Oct 26, 2026

More from juliusbrussee/caveman

All 25
1inference-sh avatarcompetitor-teardowninference-sh/skillsStructured competitive analysis with feature matrices, SWOT, positioning maps, and UX review. Covers research frameworks, pricing comparison, review mining, and107Kskills.sh installs
2firecrawl avatarfirecrawlfirecrawl/cliSearch, scrape, and interact with the web via the Firecrawl CLI. Use this skill whenever the user wants to search the web, find articles, research a topic, look104.2Kskills.sh installs
3emilkowalski avatarbreak-uiemilkowalski/skillsTry to break a piece of UI by feeding it worst-case data — long names, unbreakable emails, one-letter names, missing fields, huge counts, zero items, long label100.1Kskills.sh installs
4cloudflare avatarsecurity-auditcloudflare/security-audit-skillSecurity audit of a codebase — web apps, APIs, services, CLI tools, libraries, daemons, and more. Use when asked to find security bugs, do a security review, au95.6Kskills.sh installs
5coreyhaines31 avatardirectory-submissionscoreyhaines31/marketingskillsWhen the user wants to submit their product to startup, SaaS, AI, agent, MCP, no-code, or review directories for backlinks, domain rating, and discovery. Also u93.2Kskills.sh installs
6coreyhaines31 avatarschemacoreyhaines31/marketingskillsWhen the user wants to add, fix, or optimize schema markup and structured data on their site. Also use when the user mentions "schema markup," "structured data,87.1Kskills.sh installs