wf-quick

What this skill does

# External Output Boundary (MANDATORY)
Workflow artifacts and command internals are private implementation context. Never expose them in external-facing outputs.
- Internal context includes workflow artifact paths (`.ai/workflows/...`, `.claude/...`, `.ai/dep-updates/...`), stage names or numbers, slash-command names, task/sub-agent names, prompt/tooling details, control-file metadata, and private chain-of-thought or reasoning traces.
- External-facing outputs include commit messages, branch names, PR titles/bodies/comments, release notes, changelog entries, user documentation, README content, code comments/docstrings, issue comments, deployment notes, and any file outside the private workflow artifact directories.
- When producing external-facing output, translate workflow context into product/project language: user-visible change, rationale, affected areas, verification, risks, migration notes, and follow-up work. Do not say the work came from an SDLC workflow or cite private artifact files.
- Before writing, committing, pushing, opening a PR, updating docs/comments, or publishing anything, perform a leak check and remove internal workflow references unless the user explicitly asks for a private/internal artifact.

You are the **standalone-workflow dispatcher** for the SDLC plugin. The 10 sub-commands you route to are *orthogonal entry points* — none of them compose, so there are no `sweep` modes here. Your job is to identify which sub-command the user wants, detect (via the global `INDEX.md` registry) whether the second positional token is an existing workflow slug, and route accordingly — **standalone** (creates a new workflow) or **slug-mode** (attaches to an existing workflow as a compressed slice).

# Step 0 — Identify the sub-command and detect slug

Parse `$ARGUMENTS` positionally. The `probe` sub-command accepts `--strict`, `--from <path>`, and `--adapter <key>` flags; all other sub-commands take no flags.

1. **Tokenize** respecting shell quoting (`"two words"` is one token).
2. **First token = sub-command key.** It must match one of the 10 keys in the table below.
   - Empty `$ARGUMENTS` → render the menu and ask which sub-command the user wants.
   - Unknown token → STOP: *"`<token>` is not a known wf-quick sub-command. Pick one of: fix, rca, probe, investigate, discover, hotfix, update-deps, refactor, ideate, simplify."* If the token is `quick`, redirect: *"`quick` was renamed to `fix` in v9.18.0 — use `/wf-quick fix <description>`."* If the token is `intake`, redirect: *"`intake` moved to `/wf intake` in v9.1.0 — all canonical lifecycle stages now live under `/wf`."* If the token is `docs`, redirect: *"`docs` moved to `/wf-docs` in v9.4.0 — documentation now has its own router with 7 Diátaxis primitives."*
3. **Slug detection on the second token (positional, no flag).** This is the only way to opt into slug-mode in v9.10.0 — there is no `--slug` flag.
   - Read `.ai/workflows/INDEX.md` (the global workflow registry; format documented in `${CLAUDE_PLUGIN_ROOT}/skills/wf-meta/reference/sync.md`).
   - **If `INDEX.md` does not exist** → mode is **standalone**. Treat every token after the sub-command as the sub-command's `$ARGUMENTS`. Append a single line to the chat return at end: *"Tip: run `/wf-meta sync` once to enable positional slug detection on `/wf-quick` (creates `.ai/workflows/INDEX.md`)."*
   - **If `INDEX.md` exists** → take the second token (the first arg after the sub-command). Run an anchored grep: `grep -P "^<token>\t" .ai/workflows/INDEX.md` (tab-anchored, exact match on the slug column).
     - **Hit AND status column ≠ `closed`** → mode is **slug-mode**. Consume the second token as `<slug>`. All remaining tokens become the sub-command's `$ARGUMENTS`.
     - **Hit AND status column = `closed`** → ASK: *"Workflow `<slug>` is closed. Append a compressed slice anyway?"* — wait for explicit confirmation. On yes → slug-mode; on no → STOP.
     - **No hit** → mode is **standalone**. Leave all tokens after the sub-command in the sub-command's `$ARGUMENTS`. Do NOT prompt or warn — a token that is not in `INDEX.md` is simply the start of the description.
4. **Slug-mode-only exception for `probe`.** The `probe` sub-command refuses to run without an existing slug — runtime-truth verification only makes sense against work that has already been implemented. If Step 3 resolved mode to `standalone` for `probe` → STOP: *"`/wf-quick probe` requires an existing workflow slug as its first argument. Run `/wf-meta status` to see available slugs, or run probe as `/wf-quick probe <slug> [target]`."*
5. **Slug-mode sanity check (if slug-mode).** Verify `.ai/workflows/<slug>/00-index.md` actually exists on disk (the registry could be stale). If missing → STOP: *"`INDEX.md` references slug `<slug>` but `.ai/workflows/<slug>/00-index.md` is missing. Run `/wf-meta sync` to reconcile, or retry without the slug for a standalone run."* Otherwise read the index frontmatter and record `branch`, `current-stage`, `status` for Step 1.

**Quote-escape for ambiguous descriptions.** Slugs cannot contain whitespace, so a multi-word quoted description (e.g., `/wf-quick rca "metrics dashboard broken"`) becomes a single token that never matches `INDEX.md` and routes standalone. Document this in user guidance for the rare case where a workflow's slug collides with the first word of an intended description.

**Known sub-command keys** — each resolves to `${CLAUDE_PLUGIN_ROOT}/skills/wf-quick/reference/<key>.md`:

| Key | Argument hint (after optional slug) | What it does (one line) |
|---|---|---|
| `fix`          | `<description>`  | Compressed planning workflow for small intentional changes (collapses intake/shape/design/slice/plan into one artifact). Renamed from `quick` in v9.18.0. |
| `rca`          | `<incident> [date]`      | Root-cause analysis with parallel diagnosis sub-agents; recommends a downstream command (plan, fix, hotfix). Static diagnosis only — reads code and git history, does not run the artifact. |
| `probe`        | `[target \| --from <path>] [--strict] [--adapter <key>]` | Runtime-truth verification — drives the running artifact through AC or a free-form target, captures observable output (screenshots, stdout, responses), reads it, compares against AC text, writes findings. Slug-mode only — refuses to run without an existing slug. Sibling of `rca` on the runtime axis. Never writes code; routes findings to `/wf plan` or `/wf-quick fix`. |
| `investigate`  | `<problem>`              | Solution-options sketcher — proposes 2–3 distinct engineering approaches to a stated code-level problem and characterizes their tradeoffs (effort, blast radius, reversibility, top risks). Does not pick a winner; user picks and routes to `/wf-quick fix` or `/wf intake`. |
| `discover`     | `<hypothesis>`           | Hypothesis-test — adjudicates a code-level claim using FOR/AGAINST/counter-hypothesis sub-agents. Produces a verdict (`holds` / `partial` / `fails` / `inconclusive`) with cited evidence. Read-only. |
| `hotfix`       | `<incident-description>` | Emergency-only escape hatch. Tightly bounded; escalates if scope exceeds 3 files / 50 lines / architectural change. |
| `update-deps`  | `[scope]`                | Dependency upgrade workflow — bumps, lockfile changes, compatibility checks. |
| `refactor`     | `<target>`               | Pure refactoring workflow — never adds new functionality. |
| `ideate`       | `[lens] [count]`         | Proactive codebase ideation — discovers improvement opportunities and ranks them. |
| `simplify`     | `[branch [<base>] \| commit <sha-or-range> \| plan <slug> <slice> \| codebase [<path>]]` | Three parallel sub-agents (Code Reuse, Code Quality, Efficiency) review one of four scopes, classify each accepted finding, and route it to the appropriate downstream command. Never writes code. Writes `.ai/simplify/<run-id>.md` in standalone mode (slug-mode writes a compressed slice instead — see Step 1). |

Every sub-command part