CLI Reference¶
Global safety flags¶
| Flag | Meaning |
|---|---|
--dry-run | Preview; default for modifying commands. |
--apply | Permit local changes. |
--yes | Confirm an external or destructive managed action. |
--non-interactive | Disable prompts; required approvals must be supplied explicitly. |
--verbose | Include detection and planning details. |
--merge skip\|merge\|replace | Requested collision policy; all collisions remain skip-only. |
--skip <items> | Comma-separated modules or paths to omit. |
Command catalog¶
| Command | Arguments and common flags | Result |
|---|---|---|
doctor | --verbose, --apply | Read-only environment and project diagnostics by default; --apply records health and audit logs. |
status | none | Registry and project status. |
validate | none | Bootstrap and managed-state validation. |
version | none | Platform and schema versions. |
create <name> | --template, --provider, --output, safety flags | New additive project plan/application. |
init | safety flags | Initialize current repository additively. |
add <module> | safety flags | Add missing module-owned assets. |
remove <module> | none | Reports unsupported; never deletes project files. |
update | safety flags | Add newly managed missing assets. |
rollback [operation] | --apply --yes | Remove unchanged files owned by the selected operation. |
templates | none | List 14 template IDs. |
org <action> | trust/import/validate/evaluate/audit/compliance | Manage and enforce signed local organization policies. |
diagnostics <action> | health/bundle/metrics/traces/profile | Inspect local telemetry or create a secret-safe bundle. |
supply-chain <action> | verify/scan/generate/artifacts | Validate controls or generate SBOM, provenance, and checksums. |
skills <action> | list/trust/verify/install/status/remove | Manage signed, policy-approved prompts and engineering skills. |
workflows <action> | validate/plan/run/resume/status | Run deterministic resumable DAG workflows with consent checkpoints. |
index <action> | build/status/query/recommend | Build, query, and produce read-only recommendations from a local metadata-only project index. |
semantic <action> | configure/status/plan/build/query | Build and query an optional, consent-gated provider embedding index. |
copilot <action> | plan/run, objective and provider flags | Produce read-only engineering plans from approved project metadata. |
capabilities | none | Map capabilities to integrations. |
integrations <action> [name] | list/status/doctor/install/init/update/remove/validate/health | Plan, initialize, validate, or update registry lifecycle. |
install [tool] | consent flags | Official global tool installation plan/action. |
reference [name] | consent flags | Reference repository clone plan/action. |
providers <action> [name] | see provider actions below | Provider lifecycle, policy, auth, invocation, dashboard. |
credentials <action> [name] | --storage local\|encrypted | Secret initialization, masked setup, rotation, validation, backup, removal. |
mcp <action> [name] | definition/host flags | Register, validate, activate, health-check, or remove. |
plugins <action> [source-or-id] | trust and safety flags | Install/update declarative signed plugins and control activation. |
cloud list | none | List supported cloud adapters. |
cloud <provider> <action> | consent flags | Prepare, validate, verify, deploy, status, health, credentials, rollback. |
cloud render <action> | Render configuration flags | Blueprint-specific lifecycle. |
docker <plan\|validate\|up\|down> | consent flags | Validate or execute Compose actions. |
dashboard | --port <port> | Start authenticated loopback dashboard. |
config [key value] | export\|import, safety flags | Read/set safe config or transfer it. |
upgrade [rollback] | safety flags | Upgrade managed workspace state or restore snapshot. |
Provider actions are list, init, configure, status, doctor, validate, update, remove, models, project, mcp, invoke, test, verify, login, logout, limits, and dashboard.
Output and exit codes¶
Commands return JSON by default; verbose diagnostics go to the command result without exposing secrets. Exit 0 means success. Exit 1 means validation, policy, missing prerequisite, rejected consent, provider, or external command failure. Unknown commands print help.
Error examples¶
Unsupported template— choose an ID fromtemplates.Initialize the project— runinit --dry-run, theninit --apply.Missing ... API key— runcredentials configure <provider> --apply.Existing file ... skipped— expected additive-only behavior.Explicit consent required— review the plan and rerun with--apply --yes.
See Getting started, providers, and operations.