caveman-discover

An agent skill by JuliusBrussee, from JuliusBrussee/caveman. Tags: analytics, llm, multi-agent, workflow.

What it does

Find and label every LLM workflow in the repository so Caveman Cloud groups spend by workflow instead of one bucket. Use for "discover workflows" or breaking LLM spend down by workflow.

Install

With the skills CLI, which installs into Claude Code, Codex, Cursor and other agents:

npx skills add JuliusBrussee/caveman --skill caveman-discover

Or copy the skill folder into Claude Code's skills directory by hand (~/.claude/skills for every project, or .claude/skills inside one):

git clone --depth 1 https://github.com/JuliusBrussee/caveman
cp -r caveman/skills/caveman-discover ~/.claude/skills/caveman-discover

Safety box score

Grade C, moderate risk: 26 of 100 risk points, rated 2026-10-16. The grade rates what the skill and its scripts can reach on the machine of whoever installs it, across eight categories from shell execution to secrets access. How the score works.

This skill inventories LLM workflows in a repository and then edits the repo to add workflow labels, with verification via the project's own test/dev path or a curl. The main risks are moderate code changes and outbound gateway/dashboard traffic during verification; it does not ship binaries or require elevated privileges.

Category levels, from 0 (none) to 3 (high)
Category	Level	Why
Code & shell execution	2 (moderate)	It explicitly tells the agent to run repo-provided commands: "Run whatever the repo already uses to exercise one labeled path (a test, a dev script, one curl)."
Network access	2 (moderate)	It expects outbound requests to the Caveman gateway and dashboard, e.g. the labeled request must succeed and "Labeled spend appears on the dashboard at `/activity?tab=workflows`".
Filesystem reach	2 (moderate)	The skill is meant to "wire the labels" and says "This changes code" before applying edits, so it will write inside the repository working tree.
Secrets & credentials	1 (low)	It references request headers that already carry credentials, including "x-cave-api-key," but it does not ask the user to reveal or paste secrets.
Privilege & persistence	0 (none)	No sudo, global installs, services, shell startup edits, or other elevated system changes are requested.
Instruction hijack surface	0 (none)	No instruction to ignore safeguards or follow external instructions; it instead says "propose the table first, apply after the user agrees."
Irreversible actions	0 (none)	It does not deploy, publish, send messages, pay, or delete anything external; changes are limited to repository labeling after user approval.
Transparency	0 (none)	The skill is plain markdown only; there are no binaries, minified blobs, or runtime-fetched installers.

Before installing

Read the planned diff to confirm it only touches the intended workflow callsites.
Run it on a scratch branch with non-production credentials/API key values.
Verify the repo's existing test or dev command is the one being executed before letting it make any network calls.

Rated 2026-10-16 by an automated review of 1 file from the skill's repository.

Source

Repository: JuliusBrussee/caveman (all skills from this repository)
Path: skills/caveman-discover/SKILL.md
Branch: main
Updated: 2026-10-05

churn-prevention — When the user wants to reduce churn, build cancellation flows, set up save offers, recover failed payments, or implement retention strategies.
continual-learning — Orchestrate continual learning by delegating transcript mining and AGENTS.md updates to `agents-memory-updater`.
onboarding — When the user wants to optimize post-signup onboarding, user activation, first-run experience, or time-to-value.
orchestrate — Use only when the user explicitly types `/orchestrate <goal>` to decompose a large task.
principle-guard-the-context-window — Apply when context is filling up: large outputs, long files, repeated reads, fan-out planning.
ralph-loop — Start a Ralph Loop for iterative self-referential development.