Caveman Evidence Review

Caveman Cloud is a hosted layer that tracks LLM spend, latency, and routing, and this skill is the read-only way to ask it what it found. Point it at a question like "where is our LLM spend going?" and it pulls cost, workflow, and trace data for a bounded time window, then tests the leading explanation against a cohort of traces before making any claim. It keeps measured list-price cost, inferred daily headroom, and verified ledger savings in separate buckets so numbers never get blended together, and it refuses to pull prompt or completion payloads unless you explicitly ask. What I like is the discipline: an empty result is reported as "no current signal" rather than zero, and every finding cites trace ids. The tradeoff is that it only reads and never acts, so you still need a separate lifecycle tool to apply changes.

skills.sh installs: 119K
GitHub stars: 118.7K
last updated: Oct 26, 2026

View on GitHub

Install

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

Installs into .claude/skills of the current project.

Files

Review Caveman evidence

Act as a read-only operator. Build conclusions from current Caveman data, not from repository guesses. Never start, approve, cancel, or roll back an experiment from this skill.

Hard rules

Keep these buckets separate:
- measured provider-complete list-price cost;
- inferred daily headroom;
- verified ledger savings;
- evidence cost. Never add or relabel them.
Do not fetch prompt, completion, tool, or artifact payloads unless the user explicitly asks for payload review. Metadata, spans, timing, models, token counts, status, and optimizer attribution are enough for the default review.
Scope every read to the project selected by Caveman context. Never supply an organization id.
Empty results are evidence of no current signal, not zero cost or zero risk.
Cite trace ids and exact time windows used. Do not claim a cause from an aggregate alone.

Step 1 — Load context

Prefer MCP:

caveman_context {}

CLI fallback:

caveman cloud whoami
caveman cloud projects list

Stop if login or project selection is missing. Ask the user to run caveman login or select a project; never guess.

SKILL.md on GitHub

Details

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