@ai-squad/cli
v0.7.2
Published
ai-squad CLI — deploy SDD/Discovery squads. Skills+agents user-global (~/.claude/), hooks per-repo (<repo>/.claude/hooks/) to keep agentops telemetry consistent with hook version.
Maintainers
Readme
@ai-squad/cli
CLI to deploy ai-squad squads into Claude Code. Skills + agents install user-globally; hooks install per-repo (so the .agent-session/ telemetry your @ai-squad/agentops reports rely on stays consistent with the hook version that produced it).
Install
npm i -g @ai-squad/cliUsage
# Run once per consumer repo — installs skills+agents globally AND hooks locally.
cd <consumer-repo>
ai-squad deploy
# Selective:
ai-squad deploy --squad sdd # only the sdd squad
ai-squad deploy --hooks-only # re-sync just hooks (after upgrading the CLI)
ai-squad deploy --global-only # skip per-repo hooks (CI / dotfile flow)
ai-squad deploy --cursor # also mirror to ~/.cursor/
ai-squad helpAfter deploy:
- Skills land flat in
~/.claude/skills/<skill>/(user-global) - Subagents in
~/.claude/agents/<agent>.md(user-global) - Python hooks in
<repo>/.claude/hooks/<name>.py(per-repo, chmod +x) .claude/hooks/is appended to the repo's.gitignoreautomatically
Hook scripts are referenced from component frontmatter as
python3 $CLAUDE_PROJECT_DIR/.claude/hooks/<name>.py — Claude Code expands
$CLAUDE_PROJECT_DIR to the current project root before invoking the hook.
Why per-repo hooks?
- Telemetry consistency: the hooks that produced
.agent-session/<id>/dispatch-manifest.jsonship in the same repo, so the report engine and the data always agree on schema. - Reproducible CI: runners don't need a pre-seeded
$HOME—ai-squad deployin setup works. - Multi-version safe: different consumer repos can run different ai-squad versions concurrently without
~/.claude/hooks/collisions. - Inspectable:
cat <repo>/.claude/hooks/verify-tier-calibration.pyshows exactly what is running, no guessing about which~/.claude/version got installed.
Upgrading
npm i -g @ai-squad/cli@latest
cd <each consumer repo>
ai-squad deploy --hooks-only # pick up the new hooksThe global skills+agents update automatically on any ai-squad deploy.
Squads bundled
- sdd — Spec-Driven Development pipeline (4 phases: Specify → Plan → Tasks → Implementation). Subagents:
dev,code-reviewer,logic-reviewer,qa,blocker-specialist,audit-agent. - discovery — Cagan-style Discovery (Frame → Investigate → Synthesize). Subagents:
codebase-mapper,risk-analyst.
Companion: observability
For session reports (tokens, cost, AC closure, reviewer findings):
npx @ai-squad/agentops reportRequirements
- Node ≥ 18
- Python 3.8+ on PATH (hook scripts are pure-stdlib Python)
License
MIT — Gabriel Andrade
