Command Glossary — every /wb*, every wb-flow flag

Every /wb* command, every wb-flow subcommand and flag, and what each looks like from bash, Python and Node. Rows with no equivalent are listed anyway and say so — an absent alternative is information.

Adapted from the internal model_reference_manual.md. That manual also carries the per-provider model rosters and subscription inventory, which are environment-specific and stay internal; this page is the part that describes the shipped surface.

The distinction that runs through this whole table

28 of the 33 /wb* commands are agent-only. 5 have a deterministic CLI behind them.

	Agent-only	CLI-backed
What it is	a prompt an agent reads and acts on	code with tests
Reproducible?	no — same input can yield different output	yes — same input, same output
Callable from Python/Node?	only by dispatching an agent (see the universal shape below)	directly
Fails how?	silently, in prose	non-zero exit

This is why wb-flow model, wave, next and snap exist at all: each replaced a step a template used to ask an agent to improvise, and improvisation is where the drift came from — a hand-rolled ln -s, a mis-escaped \|\|, a slug that never existed.

The universal programmatic shape

Any /wb* command — agent-only included — can be invoked programmatically by dispatching it to an agent CLI. This is the shape referenced as “dispatch” throughout the table:

wbRun opencode run -m <slug> --dangerously-skip-permissions --command wbAudit "<target>"
wbRun agy --model <name> --dangerously-skip-permissions -p "/wbAudit <target>"

subprocess.run(["opencode", "run", "-m", slug, "--dangerously-skip-permissions",
                "--command", "wbAudit", target], capture_output=True, text=True, timeout=600)

spawnSync("opencode", ["run", "-m", slug, "--dangerously-skip-permissions",
          "--command", "wbAudit", target], { encoding: "utf8", timeout: 600000 });

agy differs: it has no --command flag — the slash-command rides inside the prompt string, and --dangerously-skip-permissions is mandatory or headless mode auto-denies every tool and returns empty output with exit 0. See the agy note below.

1 · The 33 /wb* commands

/wb* command	What it does	Deterministic CLI equivalent
/wbActOn	Turns a diagnostic document (audit, plan, review) into a ranked executi…	— agent-only
/wbAudit	Performs a deep, honest, code-based technical audit.	— agent-only
/wbBroadcast	Generates an Announcement Kit and manages version lifecycle status.	— agent-only
/wbCheck	Pre-Flight Context Quiz to verify worker model understanding.	— agent-only
/wbClean	Scans for dead code, unused files, and obsolete dependencies.	— agent-only
/wbContext	Generates a comprehensive Context Report containing identity, dependenc…	— agent-only
/wbDebug	Isolates and fixes errors following the Scientific Method.	— agent-only
/wbDeploy	Orchestrates deployment of consumer apps to web hosts (Vercel, AWS, VPS).	— agent-only
/wbDoc	Injects JSDoc/TSDoc comments and generates README files without modifyi…	— agent-only
/wbExplain	Generate persistent, formatted explanations for a specific task ID, a c…	— agent-only
/wbGit	Analyzes current file changes and generates Conventional Commits.	— agent-only
/wbHelp	Meta-command that catalogs all /wb* commands (no args) or shows the per…	wb-flow --help · wb-flow --list
/wbIdea	Tracks, scores, and validates ideas before promoting them to active plans.	— agent-only
/wbLicense	Enforces business logic by injecting premium gating and license checks.	— agent-only
/wbModel	Dynamically updates or displays the active model selections (Planner, V…	wb-flow model — --detect --reset --pick --probe --set --json
/wbMonetize	Bootstraps or maintains the Free/Pro/Dev tier splitting plumbing.	— agent-only
/wbNext	Analyzes the current project state (recent reports, git status, stale T…	wb-flow next <plan> — --embed --json
/wbPlan	Creates a structured multi-agent task plan for a given problem or objec…	— agent-only
/wbPublish	Executes the final build and NPM deployment based on the wbRelease plan.	— agent-only
/wbRefactor	Restructures and optimizes target code without changing visual behavior…	— agent-only
/wbRelease	Orchestrates the release by resolving circular dependencies and updatin…	— agent-only
/wbReview	Performs a formal quality review of an executed task against an origina…	— agent-only
/wbSecure	Scans for vulnerabilities (secrets, XSS, insecure deps).	— agent-only
/wbSetup	Initializes the Agentic Brain for the target folder by establishing its…	wb-flow init — --scope --agents --yes (wiring only)
/wbStandup	Consolidates unfinished work (open tickets, failing audits) into a sing…	— agent-only
/wbStopTrack	Stop session tracking, write final §END section, extract derivative files.	— agent-only
/wbTest	Executes dynamic tests (UNIT, E2E, MANUAL, PERF) to verify code behavior.	— agent-only
/wbToWBC	Rewrites legacy bulky HTML/Vuetify into proprietary WBC-UI2 JSON compon…	— agent-only
/wbTrack	Toggles session tracking ON so every /wb* command produces its normal r…	— agent-only
/wbTranslate	Extracts hardcoded strings and generates localized JSON files (EN, FR,…	— agent-only
/wbValid	Validates a worker's output by running an adversarial review — a second…	— agent-only
/wbVision	Proposes 3 highly innovative, strategic features to build next.	— agent-only
/wbWork	Executes one or more tasks from a plan file as the Worker model, produc…	wb-flow wave <plan> --wave=<L> (only the --wave path)

Everything marked agent-only takes the dispatch shape above; substitute its own name for wbAudit. Python and Node differ only in subprocess.run vs spawnSync — see the Python/Node shapes above.

2 · The wb-flow CLI surface

Subcommand	Flags	bash	Python / Node
(none)	--force --dry-run --list --version --help	wb-flow	subprocess.run(["wb-flow"]) / spawnSync("wb-flow", [])
init	--scope= --agents= --templates= --yes --force --dry-run	wb-flow init --scope=project --agents=detected --yes	same argv as a list/array
model	--detect --reset --pick/-i --probe --set <role>=<slug> --timeout= --file= --json --dry-run	wb-flow model --detect --probe	subprocess.run(["wb-flow","model","--detect","--json"]) → parse stdout
next	--embed --json	wb-flow next <plan.md> --embed	json.loads(...) / JSON.parse(...) on --json
snap	--label= --copy --root= --list --json --dry-run	wb-flow snap <path> --label=x --copy	same argv; --json returns {snap,target,kind,label}
wave	--wave=<L>[:<R>] --validate --summary / --no-summary --sessions --jobs= --out= --print --list · model overrides --model= --planner= --worker= --validator= --mechanical= (and --worker-model= --mech-model= --validator-model=)	wb-flow wave <plan> --wave=A --summary	--print to stdout, then run the script yourself. --list dry-runs the routing and prints the roster file + per-role chains — authoritative over model --show

Short flags on wave take =, never a space. -W= -M= -p= -v= -w= -m= are the only accepted forms — -w=zai/glm-4.7 works, -w zai/glm-4.7 does not. A space-separated short flag doesn't error: -w is dropped and the slug falls into positionals, where it is taken as the plan path. Note also that -w= is --worker=; --wave= is -W=.

-i is --pick (on model), -n is --dry-run (on snap and init), and -M= is --model=.

3 · Universal /wb* flags

Flag	Applies to	Deterministic backing
--snap / --snap-copy	every command writing under .wb/workflows/reports/	wb-flow snap <path> [--copy]
--next	every command	wb-flow next <plan.md>
--help / -h	every command	wb-flow <sub> --help
--wave=<L>[:<R>]	/wbWork	wb-flow wave <plan> --wave=<L>[:<R>]
--summary / --no-summary · --sessions	/wbWork --wave only	pass-through to wb-flow wave
-p= -v= -w= -m= (per-role model) · -M=	dispatching commands	wb-flow wave … -w=<slug>; persist with wb-flow model --set <role>=<slug>
--id --p --est --done --valid (column filters)	plan-table commands	— agent-only
--open --def --can (state overrides)	/wbWork /wbValid /wbPlan	— agent-only
--as="<styles>"	/wbWork /wbExplain	— agent-only
--yes / -y	most	--yes on wb-flow init / model
--act · --wbPlan (chaining)	/wbAudit /wbReview /wbStandup /wbActOn	— agent-only
--no-plan-update	/wbWork /wbValid	set automatically by wb-flow wave

4 · Discovery commands (not wb-flow, but you need them)

Task	bash	Python	Node
List opencode models	opencode models	subprocess.run([...]).stdout.splitlines()	execFileSync(...).split("\n")
Which providers are credentialed	opencode providers list	same, parse the ● lines	same
List agy models	agy models	same	same
List codex models	— none. codex cannot enumerate; use wb-flow model --detect --probe	—	—
Sessions as JSON	opencode session list --format json	json.loads(...)	JSON.parse(...)
Token/cost stats	opencode stats --days 1 --models	dispatch + parse	dispatch + parse