hack-cli

name	hack-cli
description	Use the hack CLI for local runtime orchestration (compose, DNS/TLS, logs, env, persistent project workspaces) and agent setup. Trigger when asked to run/start/stop services, inspect logs, manage lifecycle/workspace workflows, or update agent integrations. Prefer CLI over MCP when shell access is available.

hack CLI

Use hack as the primary interface for local-first development.

Integration freshness

• These instructions were generated by hack CLI v3.4.1; treat cached rules from another version as potentially stale.
• At session start, audit project and global integrations with hack setup sync --all-scopes --check.
• If anything is stale, missing, or deprecated, run hack setup sync --all-scopes, then reload the agent session so cached instructions are replaced.
• Never copy or hand-edit generated Hack rules to refresh them; update the CLI and run the sync command.
• Content revision: 902b04fe1469 (version alone is not a freshness guarantee).

Product boundary

• Supported v3 surface: project init, up/down/restart, open, logs, env, host exec/shell, sessions, doctor, and daemon.
• Removed surfaces: hosted auth/account/org/team flows, web dashboard, built-in GitHub workflows, and built-in Linear sync.
• Experimental and unsupported: remote/gateway/node/dispatch commands. They are hidden from default help (list with hack help --all) and warn on use; do not use them unless explicitly requested.

Operating rules

• Prefer hack over raw docker / docker compose for project workflows.
• Do not start/stop services from Docker Desktop UI for hack-managed projects.
• Treat .hack/.internal and .hack/.branch as hack-managed artifacts; do not hand-edit generated files there.
• Use --json for machine-readable output when available; hack up/down/restart/doctor --json emit an {ok, data | error: {code, message}} envelope with stable E_* error codes.
• Scripted/agent runs: pass --no-interactive (or set HACK_NO_INTERACTIVE=1) so commands never block on prompts — they apply documented defaults or fail fast with E_INTERACTIVE_REQUIRED.
• Use MCP only when shell access is unavailable.
• If runtime state looks wrong, run hack doctor, then hack doctor --fix before manual repair.

Core objects

• Project: a repo with .hack/ config + compose file.
• Service: a compose service (e.g. api, web, worker).
• Instance: a running project; branch instances are separate copies started with --branch.

Config + schema

• Project config: .hack/hack.config.json
• Global config: ~/.hack/hack.config.json
• Schema URL: https://schemas.hack/hack.config.schema.json
• Prefer CLI writes: hack config get <path>, hack config set <path> <value>, hack config set --global <path> <value>

Hostname routing + Caddy labels

• Primary host comes from dev_host (default: <project>.hack).
• Subdomain pattern is <sub>.<dev_host> (for example: api.myapp.hack).
• OAuth alias (when enabled) also routes <dev_host>.<tld> and <sub>.<dev_host>.<tld> (default tld: gy).
• Browser launches automatically prefer a routed OAuth alias when enabled; custom development hosts outside .hack stay on the primary host. Set open.prefer or pass hack open --prefer <auto|alias|dev> to override.
• Not every compose service is routable: only services with Caddy labels and on hack-dev are exposed.
• Required labels for HTTP services: caddy, caddy.reverse_proxy, caddy.tls=internal.
• Quick checks: hack open, hack open <sub>, hack open --json.

TLS + valid-hostname constraints

• hack uses Caddy internal PKI for HTTPS on routed hosts; trust CA with hack global trust.
• Containers get a combined public+local trust bundle (SSL_CERT_FILE etc.) once hack global trust has run; public TLS (package registries, external APIs) keeps working alongside *.hack trust.
• If the combined bundle is missing, only Node gets *.hack trust (NODE_EXTRA_CA_CERTS); OpenSSL-based tools keep public roots — run hack global trust to enable both.
• .hack is local-first and great for dev, but it is not a public suffix.
• Use OAuth alias hosts (for example *.hack.gy) when providers require public-suffix-style callback domains.
• Alias hosts are still local-dev routes unless you add an external tunnel/remote ingress path.

Project files (managed vs generated)

• Source-of-truth files: .hack/docker-compose.yml, .hack/hack.config.json, .hack/hack.env.default.yaml, and optional .hack/hack.env.<overlay>.yaml.
• Worktree-local env override files: .hack/hack.env.local.yaml and .hack/hack.env.<overlay>.local.yaml.
• Local-only files: .hack.secret.key, optional .hack/.env compatibility output, .hack/.env.state.json, and .hack/.internal/ (runtime/local machine state; keep gitignored).
• Generated (do not hand-edit): .hack/.internal/compose.override.yml, .hack/.internal/compose.env.override.yml, .hack/.internal/compose.runtime.override.yml, .hack/.branch/compose.<branch>.override.yml, .hack/.branch/compose.<branch>.runtime.override.yml.
• Managed via CLI: .hack/.internal/extra-hosts.json (use hack internal extra-hosts ... commands).
• Lifecycle runtime files: .hack/.internal/lifecycle/state.json, .hack/.internal/lifecycle/*.log.
• Ignore rules: hack owns a committed .hack/.gitignore (self-healing on init/up) covering machine-local generated files (.internal/, .branch/, .env, .env.state.json, hack.env*.local.yaml); keep it committed, and if generated files leaked into git, hack doctor --fix untracks them (files stay on disk).

Linked git worktrees

• Secret key inherits from the primary checkout automatically through the shared git common dir; set HACK_ENV_SECRET_KEY for CI or detached environments.
• hack up in a linked worktree defaults to a branch instance named after the worktree's git branch; a detached linked worktree requires an explicit --branch, unless config worktree.auto_branch=false explicitly opts into the base instance.
• Before hack up or hack restart auto-targets a new branch instance, Hack warns when the same worktree already owns a non-terminal instance; pass --branch <name> to make the target explicit.
• hack doctor flags divergent secret keys and dev_host collisions across checkouts.

Advanced networking (extra_hosts + local proxies/tunnels)

• Static host mappings: set internal.extra_hosts in .hack/hack.config.json.
• Dynamic host mappings: hack internal extra-hosts set <hostname> <target> / unset / list.
• For host-local proxies/tunnels, prefer host-gateway as target when possible.
• After mapping changes or proxy IP churn: hack restart and then hack doctor.

Standard workflow

• If .hack/ is missing: hack init
• Start services: hack up --detach (or hack up -d)
• Check status: hack ps or hack status
• Open app URL: hack open --json
• Restart: hack restart
• Stop services: hack down

Running things (decision guide)

• One-off command in a fresh service container (deps started as needed): hack run <service> <cmd...>.
• Command inside an already-running service container: hack exec <service> -- <cmd...>.
• Host script that needs hack-stored env: hack host exec --env <overlay> --scope <service> -- <cmd...> — this is THE way to run repo scripts; never read .env files directly.
• Interactive host shell with injected env: hack host shell --env <overlay> --scope <service>.
• Browser/host URL: use hack open <service> --json (OAuth aliases are preferred when enabled). Container-to-container traffic should use Compose DNS.

Logs (default is compose)

• Fast tail: hack logs --pretty
• Per-service tail: hack logs <service>
• Machine snapshot: hack logs --json --no-follow
• Loki history/query: hack logs --loki --since 2h --pretty or hack logs --loki --query '{project="<name>"}'
• Force compose backend: hack logs --compose
• Global infra logs: hack global logs caddy --no-follow --tail 200

Lifecycle + startup

• Put host setup in .hack/hack.config.json under startup/lifecycle (not ad-hoc terminal tabs).
• Use lifecycle.up.before for pre-start hooks and lifecycle.processes for long-running host tasks.
• Lifecycle hooks and processes receive the selected overlay's global values plus host overrides; service-scoped values remain container-specific unless a host command explicitly selects that scope.
• For fixed-port host helpers such as SSM tunnels or local proxies, set singleton.ports and usually onConflict: "adopt" so Hack reuses a healthy existing listener instead of starting duplicate tunnel stacks.
• singleton is a listener guard, not process ownership transfer; adopted external processes are left running on hack down.
• Inspect lifecycle status via hack projects --details and stream via hack logs <service-or-process>.
• Lifecycle session recovery is ownership-proven: Hack adopts healthy token-, definition-, and environment-matched sessions, replaces owned stale sessions, and refuses to kill same-name sessions without deterministic ownership proof.
• hack doctor --fix reaps an orphan lifecycle session only when mux ownership is proven and its Compose instance is absent; unverified same-name sessions are never modified.
• After hack up or hack restart, running services and successful one-shot services (exited with code 0) count as successful. Hack recognizes dependency installers, hack.service.one-shot=true, and services referenced by Compose condition: service_completed_successfully; other states return E_STARTUP_INCOMPLETE, and hack doctor warns about containers stuck in Created.
• Detached startup is bounded; a hung Compose operation returns E_STARTUP_TIMEOUT, terminates its process group, and hack doctor --fix can start exact containers left in Created.
• Target only affected services with hack up <service...> --detach, hack restart <service...>, or hack env apply --service <service>; scoped operations skip project lifecycle hooks and implicit dependency startup.
• Use hack env explain <KEY> --env <overlay> --service <service> --target <host|compose> for redacted source, precedence, availability, and delivery diagnostics.
• Dependency installer services are detected generically by command or hack.dependencies.bootstrap=true; registry env references are preflighted before container mutation. Optional hack.dependencies.cache-volume, hack.dependencies.lockfiles, and hack.dependencies.runtime-files labels enable lockfile/runtime-keyed volumes shared across compatible worktrees.

Workspaces (mux-managed, tmux-first by default)

• Picker: hack session for persistent project workspaces.
• Reuse/create: hack session start <project>
• Env-scoped workspace: hack session start <project> --env qa --service api --detach
• Force isolated agent workspace: hack session start <project> --new --name agent-1 (<project>--agent-1).
• Execute in workspace: hack session exec <workspace> "<command>"
• Execute in workspace with injected env: hack session exec <workspace> --env qa --service api "bun db:migrate"
• Stop workspace: hack session stop <workspace>

Host-side env helpers

• One-off host command with injected env: hack host exec --env qa --scope api -- bun db:migrate
• Host commands default to a host-local env view; use --target compose when you explicitly want container-oriented addresses.
• --scope selects which env scope to inject; it does not move execution into that service container.
• Interactive host shell with injected env: hack host shell --env qa --scope api
• Run inside an already-running service container: hack exec api -- bun test

Branch instances (parallel envs)

• Use a branch instance when you need two versions running at once (PR review, experiments, migrations) or want to keep a stable environment while testing another branch.
• Target one with --branch <name> on up/open/logs/down (for example: hack up --branch <name> --detach).
• Linked worktrees pick a branch instance automatically (see Linked git worktrees).
• Containers receive HACK_RUNTIME_METADATA plus HACK_DEV_URL, HACK_ALIAS_URL, and current-service URL fields derived from effective Caddy routes. Use Compose DNS for internal traffic and this metadata for browser-facing links, OAuth callbacks, and webhooks.

Run commands inside services

• One-off: hack run <service> <cmd...> (uses docker compose run --rm)
• Example: hack run api bun test
• Use --workdir <path> to change working dir inside the container.
• Use hack ps --json to list services and status.

Project targeting

• From repo root, commands use that project automatically.
• Else use --project <name> (registry) or --path <repo-root>.
• List projects: hack projects --json

Global infra

• Bootstrap once: hack global install
• Start/stop/status: hack global up, hack global down, hack global status
• Use hack global up before Loki/Grafana queries if global logging is offline.

Daemon (optional)

• Start for faster JSON status/ps: hack daemon start
• Check status: hack daemon status

Docker compose notes

• Prefer hack commands; they include the right files/networks.
• Use docker compose -f .hack/docker-compose.yml exec <service> <cmd> only if you need exec into a running container.

Agent integration maintenance

• Project-level hack commands auto-check integration drift and attempt auto-sync (project docs, client skills/rules, shared global skills, and MCP).
• When drift is detected, Hack reports it before repair and tells the agent to reload after repair; it never silently leaves the session using cached rules.
• Set HACK_SETUP_SYNC_MODE=warn to only warn, or HACK_SETUP_SYNC_MODE=off to disable.
• Refresh project + user integrations: hack setup sync --all-scopes
• Audit integration state only: hack setup sync --all-scopes --check
• Remove generated integration artifacts: hack setup sync --all-scopes --remove
• After upgrading CLI: hack update then hack setup sync --all-scopes
• When changing hack itself: interface or behavior changes must update docs/ in the same change (regenerate the CLI reference with bun run docs:cli-reference).

Agent setup (CLI-first)

• Cursor rules: hack setup cursor
• Claude hooks: hack setup claude
• Codex skill: hack setup codex
• Refresh all local agent integrations: hack setup sync --all-scopes
• Agent-assisted onboarding: hack init --with claude|codex|both (new repos) or hack agent onboard (existing projects) print/hand off the full setup prompt; the /hack-init skill and the hack-init MCP prompt return the same content.
• Init prompt: hack agent init (use --client cursor|claude|codex to open)
• Init patterns: hack agent patterns
• MCP (no-shell only): hack setup mcp
• MCP install (explicit): hack mcp install --all --scope project