Stack bootstrap

Skill drmahdikazempour/agent-stack/skills/stack-bootstrap

One command to make any repo token-efficient for Claude Code & Cursor. Built-in code map, output compression, skills & hooks, savings measured with ccusage.

Install

npx -y skills add drmahdikazempour/agent-stack --skill stack-bootstrap

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Set up or re-sync the agent-stack optimization layer (CLAUDE.md, skills, hooks, code graph, shell compression, Cursor mirror) for a repo. Use when onboarding a repo, right after cloning, or when the agent context feels bloated/unoptimized. Runs `agent-stack init` under the hood — detect, install adapters, generate files, wire hooks, verify, and capture a token baseline — then reports exactly what changed.

SKILL.md

2.0 KB, as published. Nobody here has run it

stack-bootstrap

Bootstrap or re-sync this repo's optimization layer. Skills decide when; the CLI decides how — always shell out to agent-stack for file writes so the hook merger stays the sole writer of settings.json.

When to use

First time working in a repo that should have agent-stack configured.
After pulling changes that touched .claude/, .cursor/, or integrations/.
When startup context feels heavy and you suspect the setup drifted.

What to do

Run npx @drmahdikazempour/agent-stack doctor to see current state and whether anything is configured.
If unconfigured or drifted, run npx @drmahdikazempour/agent-stack init. It:
- detects host(s), profile, package manager, and existing config,
- backs up anything it will touch into .agent-stack.bak.<ts>/,
- installs the profile's pinned adapters (or runs config-only with --no-install),
- generates CLAUDE.md, skills, agents, commands, .claudeignore, the Cursor mirror, and MCP scaffold,
- wires hooks in a single merged write,
- verifies every skill loads and every hook is present (rolls back on failure),
- captures a ccusage token baseline.
Report the chosen profile, adapters installed, hooks wired, and baseline tokens.

Flags worth knowing

--dry-run — print the plan, write nothing.
--profile <code|review|multimodal|spec> — override auto-detection.
--no-install — write configs only; skip binary installs.
--yes — skip the single confirm prompt.

Idempotency

Running init twice is a no-op unless --force. Safe to re-run anytime.