@launchsecure/launch-kit 0.0.28 → 0.0.30

package/scaffolds/ls-marketplace/plugins/kit/skills/beacon-pulse.md

@@ -0,0 +1,82 @@
+---
+description: Show the most recent failure from the active launch-beacon monitor session, plus the N events that preceded it (default 10). Answers "what was happening when it broke" without scrolling through the full stream. Read-only.
+---
+
+# /kit:beacon-pulse
+
+The "what happened just before this failure" command — a single pulse, freezing the moment the beacon caught a failure. When a bug fires, the actual story is usually the chain of events leading up to it: the click that triggered a state change, the failed fetch that left auth in a bad state, the route change that mounted the wrong component.
+
+A "failure" here is anything `mcp__local-launch-beacon__beacon_failures` returns — window errors, unhandled rejections, fetch/xhr with status ≥ 400 or thrown, and clicks where overlay interception blocked the intended target. Broader than just `kind=error`; matches what users mean when they say "something broke."
+
+Parse `$ARGUMENTS` as space-separated tokens, in any order:
+- An optional **integer** — the count of preceding events to include. Default 10. Cap at 50 (the `beacon_correlate.before` tool limit).
+- `--worktree=<slug>` — orbit worktree slug. Forwarded to every MCP call as the `worktree` arg.
+- `--project_root=<path>` — explicit project root (absolute, or relative to cwd). Forwarded as `project_root`. Takes precedence over `--worktree`.
+
+Examples:
+- `/kit:beacon-pulse` → 10 events before the latest failure
+- `/kit:beacon-pulse 25` → 25 events before
+- `/kit:beacon-pulse --worktree=beacon_rewrite` → query a worktree's session
+
+## Locate the most recent failure
+
+Call `mcp__local-launch-beacon__beacon_failures` with `limit: 1` (and `worktree` / `project_root` if the user supplied them). The tool implicitly targets the most recent session.
+
+If the response is `{error: "no session found"}` OR an empty failures array, **alert the user with the exact MCP error string**, then walk the fallback below — do NOT silently grep. The MCP can come up empty for two non-bug reasons: (a) no session has been started, (b) the session NDJSON lives outside the MCP's watch root (e.g. inside an orbit worktree the user didn't name).
+
+**Fallback when the MCP returns nothing:**
+1. If the user did NOT pass `--worktree` / `--project_root` and `launch-orbit` has worktrees registered for this repo, suggest re-running with `--worktree=<slug>` before falling back to the filesystem.
+2. Look for `.launchsecure/beacon-*.ndjson` in cwd, then in up to 2 parent dirs. Use the `Read` tool to inspect the most-recently-modified matching file — do NOT use `grep`, `find`, or `ls` as the primary path.
+3. If files are found at an alternative path, tell the user the path mismatch explicitly: `"MCP scans <root>/.launchsecure/ but the session lives at <found-path>. Reading directly."`
+4. If nothing is found anywhere, stop with: `"No beacon sessions found. Start one with: npx launch-beacon monitor"`.
+
+Capture the failure's `seq`, `kind`, `ts`, and message.
+
+## Pull the context window
+
+Call `mcp__local-launch-beacon__beacon_correlate` with `seq: <failure.seq>`, `before: N`, `after: 0`, plus the same `worktree` / `project_root` args you used on `beacon_failures`. The tool returns event summaries directly — no JSON parsing required.
+
+If a context event's summary is missing detail you need for analysis (truncated stack, missing response body, selector chain), call `mcp__local-launch-beacon__beacon_event` with that event's specific `seq` for the full JSON — and forward the same root args. Only do this when the summary is insufficient — not unconditional.
+
+## Output
+
+Render a **timeline** with the failure visually separated at the bottom:
+
+```
+session: <token>
+window:  N events before failure at seq <seq>
+
+12:34:51  click       button.submit-btn
+12:34:51  fetch       POST 401 /api/sessions (120ms)  FAILED
+12:34:51  fetch       GET 401 /api/me (45ms)  FAILED
+12:34:52  route       /login → /dashboard (pushState)
+...
+
+╴╴╴╴ FAILURE ╴╴╴╴
+12:34:53  error       Cannot read properties of undefined (reading 'orgSlug')
+              at apps/Console.tsx:42:18
+              stack: Console.render
+                     React.renderWithHooks
+                     React.beginWork
+```
+
+Format rules:
+- One line per event, same SUMMARY format as `/kit:beacon-scan`.
+- The failure block at the end gets the message + (truncated) stack frames if present. For fetch/xhr failures, include method + status + url + response body summary.
+- Separator above the failure is the visual cue.
+
+After the timeline, write **one short paragraph** of analysis: what does the chain *look* like? Common patterns:
+- 401s → error referencing user/org → "auth expired or org context missing"
+- click → no observable effect → click again → error → "first click handler crashed silently; second click landed on something else"
+- route change → error → "component on the new route blew up during initial render"
+- repeated 400s on the same endpoint → "request shape mismatch — check the server's Zod schema against the client payload"
+
+Don't dive into a fix — name the suspected proximate cause, suggest one concrete next step (which file to read, which probe to add, which `/kit:beacon-scan <kind>` to drill into).
+
+## Constraints
+
+- **MCP-first.** Use `beacon_failures`, `beacon_correlate`, `beacon_event` — never `grep`, `find`, `ls`, `sed`, `cut`, or `tail` as the primary path.
+- **Alert on fallback.** If the MCP can't answer, say so verbatim before reading the file directly.
+- **Read-only.**
+- **Most recent failure only.** For an earlier one, clear the session (`/kit:beacon-clear`) and reproduce, or use `/kit:beacon-scan error` / `/kit:beacon-scan fetch` to browse.
+- **Plain text.** Same monospace conventions as `/kit:beacon-scan`.

package/scaffolds/ls-marketplace/plugins/kit/skills/beacon-scan.md

@@ -0,0 +1,66 @@
+---
+description: Scan recent events from the active launch-beacon monitor session via the launch-beacon MCP. Defaults to last 50 events across all kinds; pass a kind (`error`, `click`, `fetch`, `route`, `dialog`, `probe`, etc.) to filter, and/or a numeric limit. Read-only.
+---
+
+# /kit:beacon-scan
+
+Recent runtime events captured by a `launch-beacon monitor` session. Use this as the first step when investigating something that just happened in the browser — clicks, failed network calls, errors, route changes, probe() calls, dialog mount/unmount.
+
+Parse `$ARGUMENTS` as space-separated tokens, in any order:
+- A **kind filter**: one of `error`, `unhandledrejection`, `fetch`, `xhr`, `click`, `mousedown`, `route`, `dialog`, `submit`, `probe`. Anything else that doesn't parse as an integer or a known `--flag` is treated as an explicit session token.
+- A **limit**: integer. Default 50. Hard cap 500 (the `beacon_recent.limit` tool ceiling).
+- An explicit **session token** (8-char hex like `8f3c2a91`) — passed straight through as the `session` MCP arg.
+- `--worktree=<slug>` — orbit worktree slug (from `launch-orbit create`). Forwarded to the MCP as the `worktree` arg so a Claude session pinned to the main repo can read sessions captured inside a worktree.
+- `--project_root=<path>` — explicit project root (absolute, or relative to cwd). Forwarded to the MCP as `project_root`. Takes precedence over `--worktree`.
+
+Examples:
+- `/kit:beacon-scan` → last 50 events of any kind from the latest session
+- `/kit:beacon-scan error` → last 50 errors only
+- `/kit:beacon-scan click 20` → last 20 click events
+- `/kit:beacon-scan 100` → last 100 events of any kind
+- `/kit:beacon-scan fetch 50 8f3c2a91` → last 50 fetches from a specific session
+- `/kit:beacon-scan fetch --worktree=beacon_rewrite` → last 50 fetches captured in that worktree
+
+## Scan
+
+Call `mcp__local-launch-beacon__beacon_recent` with `kind: <kind?>`, `limit: <limit>`, `session: <session?>`, plus `worktree: <slug?>` and `project_root: <path?>` if the user supplied them. The tool defaults to the most-recent session and returns one compact summary line per event.
+
+If the response is `{error: "no session found"}` OR an empty events array, **alert the user with the exact MCP error string**, then walk the fallback below — do NOT silently grep.
+
+**Fallback when the MCP returns nothing:**
+1. If the user did NOT pass `--worktree` / `--project_root` and the cwd is the main repo of a project that has worktrees registered with `launch-orbit`, suggest re-running with `--worktree=<slug>` before falling back to the filesystem — that's usually the actual fix and avoids a wrong-directory scan.
+2. Look for `.launchsecure/beacon-*.ndjson` in cwd, then in up to 2 parent dirs. Use the `Read` tool to inspect the most-recently-modified matching file — do NOT use `grep`, `find`, or `ls` as the primary path.
+3. If files are found at an alternative path, tell the user the path mismatch explicitly: `"MCP scans <root>/.launchsecure/ but the session lives at <found-path>. Reading directly."`
+4. If nothing is found anywhere, stop with: `"No beacon sessions found. Start one with: npx launch-beacon monitor"`.
+
+## Output
+
+Group events by **time** (most recent last — chronological matches user intuition for "what happened"). For each event, render one line:
+
+```
+HH:MM:SS  KIND     SUMMARY
+```
+
+Where SUMMARY is a one-liner pulled from the event body:
+- **error / unhandledrejection** → `message` (truncate ≤ 80 chars)
+- **fetch / xhr** → `METHOD status url (durationMs)` — flag `FAILED` if `failed:true`
+- **click / mousedown** → `target.selector` and, when `hitMatchesTarget === false`, also append `(occluded by <hitTarget.selector>)` — that's exactly the diagnostic the user is here for
+- **route** → `from → to (via)`
+- **dialog** → `<action> <selector>` (e.g. `mount dialog#xyz` or `unmount section[role=dialog]`)
+- **submit** → `<form selector> (preventDefault?)`
+- **probe** → `<label>: <truncated JSON>`
+
+If a summary line is missing detail you need (truncated response body, full selector chain, stack frames), call `mcp__local-launch-beacon__beacon_event` with that event's `seq` for the full JSON — and forward the same `worktree` / `project_root` args you used on `beacon_recent`. Only on demand — not unconditional.
+
+After the timeline, if anything looks suspicious (clusters of failed network calls, a click followed by no observable effect, a `hitMatchesTarget:false` event, an unhandled rejection), call it out in **one short paragraph** at the end. Don't analyze every event — only surface the standouts.
+
+If the buffer is empty after filtering, say so: `No events match that filter in <session>.`
+
+## Constraints
+
+- **MCP-first.** Use `beacon_recent` / `beacon_event` — never `grep`, `find`, `ls`, `tail`, `head` as the primary path.
+- **Alert on fallback.** If the MCP can't answer, say so verbatim before reading the file directly.
+- **Read-only.** Never delete, modify, or write to the NDJSON file. Wiping is `/kit:beacon-clear`.
+- **Always use the latest session by default.** Pass an explicit token as the third arg only when the user wants a specific older session.
+- **Plain text output.** Monospace alignment beats markdown tables here.
+- **Don't speculate on root cause.** Surface anomalies; let the user direct the next query (often: `/kit:beacon-pulse` for context around the latest failure).

package/scaffolds/ls-marketplace/plugins/kit/skills/blast-radius.md

@@ -0,0 +1,101 @@
+---
+description: Build a blast radius for a file/table/endpoint via launch-chart and push it to launch-deck as an interactive radial graph. Use to assess "what breaks if I change X" before refactors, schema migrations, or API contract changes. Chart-first; on chart gaps (no nodes, missing layer, empty edges) files a feedback comment to LaunchSecure so the gap is monitored.
+---
+
+# /kit:blast-radius
+
+Calculate the reverse-dependency blast radius for a target node and visualise it as an interactive radial graph in launch-deck. The graph centres on the target and rings out by hop distance across all project layers (db / api / ui / static / etc.).
+
+Parse `$ARGUMENTS`:
+- **target** (required) — a node id from the project graph (file path like `lib/permissions/types.ts`, table name like `User`, endpoint id like `POST /api/work-items`). If absent, ask the user.
+- **hops=N** — max traversal depth. Default 2.
+- **layer=<id>** — restrict to one layer when the same name exists in multiple layers (e.g. `layer=db`).
+- **--session=<id>** — deck session name. Default `blast-<target-slug>`.
+- **--mode=feature** — render with feature rings (modify / create / ripple) instead of the default structural rings (hop1 / hop2). Use when the radius represents planned change scope rather than pure structural reachability.
+
+Examples:
+- `/kit:blast-radius lib/permissions/types.ts`
+- `/kit:blast-radius User layer=db hops=3`
+- `/kit:blast-radius POST /api/work-items --mode=feature`
+
+## Preflight
+
+1. Confirm both MCPs are wired — `mcp__launch-chart__chart_server_status` and `mcp__launch-deck__server_status`. If either is missing, stop and tell the user to run `npx @launchsecure/launch-kit refresh` (or `npx launch-deck` for the deck server).
+2. Run `mcp__launch-chart__detect_project_stack` once. Cache the returned `layers` for the bug-report fallback below.
+
+## 1. Compute the blast points
+
+Call `mcp__launch-chart__blast_points` with `node_id: <target>`, `hops: <hops>`, `direction: "reverse"`, and `layer` if supplied. The tool returns affected nodes (with hop, type, layer, module) plus a summary.
+
+**Chart-gap detection** — if any of these are true, file a feedback comment per the shared `chart-gap` protocol below before stopping:
+- `error` is set
+- `nodes` is empty AND the user explicitly named a node id that exists on disk (verify by `Read`-ing the path)
+- `summary.byLayer` covers fewer layers than `detect_project_stack` reported as available
+
+## 2. Build the manifest
+
+Translate the blast_points result into the deck `blast-radius` block manifest shape. Reuse layer ids/icons/colors from `detect_project_stack` so the visual matches the project's conventions; fall back to:
+- db → `database`, `#172554`
+- api → `server`, `#1e3a5f`
+- ui → `layout-dashboard`, `#0c4a6e`
+- static → `file-text`, `#374151`
+- shared → `puzzle`, `#4338ca`
+- middleware → `shield`, `#1f2937`
+- config → `cog`, `#52525b`
+
+Rings:
+- structural mode (default): `[{id:"hop1",name:"Direct",color:"#f97316"},{id:"hop2",name:"Indirect",color:"#eab308"}]`
+- feature mode: `[{id:"modify",name:"Modify",color:"#f97316"},{id:"create",name:"Create",color:"#22c55e"},{id:"ripple",name:"Ripple",color:"#eab308"}]`
+
+Centre node: `{ name: <target>, description: "<N> direct, <M> indirect (<hops> hops)" }`.
+
+Nodes: map each blast result row to `{ id, name, layer, ring: hop===1 ? "hop1" : "hop2", path, type }`.
+
+Edges: connect the centre to every hop=1 node; connect each hop=2 node to one hop=1 ancestor from its dependency chain (read from `blast_points` `via` field if present, otherwise pick any hop=1 node in the same layer).
+
+## 3. Push to deck
+
+Call `mcp__launch-deck__deck` with `session: <session>`, `mode: "show"`, `blocks: [{ type: "blast-radius", label: <target>, manifest: <built-manifest> }]`.
+
+If the deck server isn't running, call `mcp__launch-deck__start_server` first, then retry. Report the deck URL back to the user.
+
+## Output
+
+After the push, print a terse summary:
+
+```
+blast-radius for <target> — <N> direct, <M> indirect (<hops> hops)
+  by layer: db 4, api 7, ui 12
+  pushed to deck session "<session>" — http://localhost:52829/?session=<session>
+```
+
+## Chart-gap protocol (shared)
+
+When the chart MCP returns less data than expected, file ONE feedback comment in the project's Comm Hub (cloud LS by default per project memory — use `mcp__launch-secure__communication_write`, not the local variant) before stopping:
+
+```
+resource_type: "feedback"
+title: "launch-chart gap: <one-line>"
+body: |
+  Skill: /kit:blast-radius
+  Tool:  blast_points(node_id=<id>, hops=<n>, layer=<layer?>)
+  Returned: <nodes_count> nodes, layers=<layers_seen>
+  Expected: <what-was-expected, e.g. "node exists at <path>" or "layer X should appear">
+  Project stack: <one-line summary from detect_project_stack>
+fields:
+  severity: "low"
+  source: "kit-skill"
+  skill: "blast-radius"
+  tool: "blast_points"
+  args: { node_id, hops, layer }
+  response_summary: { nodes_count, layers_seen, error }
+```
+
+Do NOT spam — file at most one comment per skill invocation. After filing, tell the user "Filed chart-gap report to LS (id: …). Falling back to read_graph + Read." and continue with a degraded path: try `mcp__launch-chart__read_graph` with `search: <target>` to confirm the node exists in any layer; if so, do the blast-radius math from `include_edges: true` results manually; otherwise stop and surface the gap to the user.
+
+## Constraints
+
+- **MCP-first.** Never grep the filesystem to build the manifest — use `blast_points` and `read_graph`.
+- **Idempotent on re-run.** Re-pushing to the same session overwrites the previous render — that's fine, just tell the user.
+- **Single bug-file per run.** Don't file a comment for every gap; bundle into one report.
+- **Plain text output.** No markdown tables, no fenced blocks in the human-facing summary.

package/scaffolds/ls-marketplace/plugins/kit/skills/brief.md

@@ -0,0 +1,112 @@
+---
+description: Create or update a repo-backed Brief in the LaunchSecure Comm Hub per the LS Briefs framework — Discovery-plane document (docCategory=features) authored as a comment that's also synced to a markdown file in the repo. Use to capture an idea before promoting it to an Epic.
+---
+
+# /kit:brief
+
+A **Brief** is an LS Discovery-plane artefact: a long-form description of a problem/opportunity, authored as a comment in the project's Comm Hub AND backed by a markdown file in the repo (`docs/requirements/features/<slug>.md` by convention). It's the structured alternative to a free-form Discussion when the idea has enough shape to write down. Briefs get promoted to Epics on the Roadmap once they're DECIDED.
+
+Parse `$ARGUMENTS`:
+- **subcommand** (required) — `new` | `update` | `list` | `show`.
+- For **new**: `<title>` (required), `--from=discussion:<id>` to seed body from an existing discussion thread, `--module=<slug>` to tag.
+- For **update**: `<brief-id-or-slug>` and either a body description (re-generates body) or `--append=<text>` to add a section.
+- For **list**: `[--module=<slug>] [--status=open|decided|declined]`.
+- For **show**: `<brief-id-or-slug>`.
+
+Examples:
+- `/kit:brief new "Channels for Comm Hub" --module=communication`
+- `/kit:brief new "Roadmap nav" --from=discussion:cmt_abc123`
+- `/kit:brief list --status=open`
+- `/kit:brief show channels-for-comms`
+- `/kit:brief update channels-for-comms --append="Resolved: use the Slack metaphor not Teams."`
+
+## Preflight
+
+1. Confirm `mcp__launch-secure__ping` works (the cloud LS MCP must be wired). If only `mcp__local-launch-secure__*` is wired, ask the user whether to post to local or cloud — default per project memory is CLOUD.
+2. For `new`, sanity-check that no existing brief with the same slug already exists — call `mcp__launch-secure__communication_read` filtered by `resourceType=comment` and search the title. If a match is found, ask whether to update instead.
+
+## The Brief shape
+
+Per the LS Briefs framework (`docs/requirements/work-items/roadmap-and-hierarchy.md`):
+
+- **resourceType**: `comment` (Briefs are comments)
+- **docCategory**: `features` (this is what file-backed sync watches)
+- **title**: the human-friendly name; the slug is derived
+- **body**: the long-form markdown — sections below
+- **file_links**: not used here; the FileBackedEntity link is created by LS server-side on first save
+- **tags**: optional module tag, plus a `brief` tag for filtering
+
+### Body template
+
+When generating a NEW brief, use this skeleton (markdown body):
+
+```markdown
+# <Title>
+
+**Status**: PROPOSED  
+**Owner**: <user-mention or "unassigned">  
+**Module**: <module slug or "-">
+
+## Problem
+
+<1-2 paragraphs — what's broken or missing, who feels it, why now>
+
+## Proposed direction
+
+<1-3 paragraphs — the shape of the solution, not the implementation>
+
+## Out of scope
+
+- <bullets — what this brief deliberately does NOT cover>
+
+## Open questions
+
+- [ ] <bullets — what we still need to decide before this can become an Epic>
+
+## Success looks like
+
+<1-2 sentences — observable outcome, not "ship X">
+
+## References
+
+- <link to existing code via launch-chart node id (use mcp__launch-chart__read_graph to find the right nodes; cite them as `lib/foo/bar.ts:42`)>
+- <related comments by id>
+```
+
+The body is plain markdown — Comm Hub renders it. Use launch-chart to add real code references (don't invent file paths).
+
+## Write paths
+
+### new
+
+1. Build the body from the template + the user's description (or the `--from=discussion:<id>` seed — fetch via `communication_read` and weave the relevant points into the Problem/Proposed-direction sections).
+2. Call `mcp__launch-secure__communication_write` with:
+   - `title: <user-supplied>`
+   - `body: <generated-markdown>`
+   - `resource_type: "comment"`
+   - `fields: { docCategory: "features", briefStatus: "PROPOSED" }`
+   - `tag_ids: <module tag id + "brief" tag id>` (call `mcp__launch-secure__tags_list` first to resolve names → ids; if either tag doesn't exist, surface the missing tag(s) and proceed without)
+3. Print the returned comment id + URL.
+4. Tell the user: "LS server-side file-backed sync will materialize `docs/requirements/features/<slug>.md` on the next default-branch push that includes this brief — pull the branch to see it locally."
+
+### update
+
+1. Resolve `<brief-id-or-slug>` via `communication_read(resourceType: "comment", search: <slug>)`. If multiple match, list them and stop.
+2. If `--append=<text>`, fetch the current body, append the text under a `## Update <ISO-date>` heading, and call `mcp__launch-secure__communication_update` with the new body.
+3. If a fresh description was passed, regenerate the body from the template + new description, ask the user to confirm (show the diff briefly), then update.
+
+### list
+
+`mcp__launch-secure__communication_read` filtered by `resourceType=comment`, narrow on `docCategory=features` if the API supports it (otherwise pull a wider set and filter client-side by tag/status). Render one line per brief: `id  status  title  module  updated`.
+
+### show
+
+Fetch via `communication_read` and print the body as-is, no transformation.
+
+## Constraints
+
+- **Plain markdown body.** Per project memory `feedback_comms_plain_text.md`, comms posts are plain — no HTML, no emojis unless the user wrote them.
+- **One brief, one comment.** Don't fragment a brief across multiple comments — append via `--append` to keep the document continuous.
+- **Cloud LS by default.** Per memory `feedback_post_to_cloud_ls_default.md`. Local LS only when the user explicitly says "local".
+- **Cite real code.** Use launch-chart node ids in the References section — don't invent file paths.
+- **Don't auto-promote to Epic.** A Brief → Epic is a deliberate human action (project memory: bucket transitions are deliberate, not derived). Surface a hint at the end: "When DECIDED, run `/work-items create epic --from-brief=<id>` to promote."

package/scaffolds/ls-marketplace/plugins/kit/skills/course.md

@@ -0,0 +1,84 @@
+---
+description: Show the currently active LaunchSecure course (server profile) and list the others available in .launch-secure.cred.config. Surfaces which serverUrl/org/project this Claude session is talking to so the user doesn't post local LS comms to prod, or vice versa. Read-only by default; switch is opt-in.
+---
+
+# /kit:course
+
+A "course" is a named LaunchSecure destination — `serverUrl + PAT + orgSlug + projectSlug` — stored under `profiles` in `.launch-secure.cred.config`. The active course drives every MCP call and every `npx launch-pod` invocation. This skill answers two questions: *which course am I on right now*, and *what others are available* — without changing anything unless explicitly told to.
+
+Parse `$ARGUMENTS` — first token is the subcommand:
+
+- (empty) or **status** — one-line: which course is active.
+- **list** — table of every course, active marked with `★`.
+- **switch <name>** — change the active course (rewrites `.launch-secure.cred.config` AND `.mcp.json`'s launch-secure URL). Asks for confirmation.
+- **diff** — show the deltas between two courses (`--a=<name> --b=<name>`).
+
+Examples:
+- `/kit:course`            → "active: prod (acme/launchsecure-v2 @ launchsecure-v2.vercel.app)"
+- `/kit:course list`       → table
+- `/kit:course switch local`
+- `/kit:course diff --a=local --b=prod`
+
+## Preflight
+
+Confirm `.launch-secure.cred.config` exists in the cwd (or in the nearest parent up to 3 levels — `cred-shape.ts`'s lookup convention). If missing, tell the user: `"No .launch-secure.cred.config here. Run \`npx @launchsecure/launch-kit init\` first."`
+
+Read the cred file directly (it's gitignored, lives at the repo root, JSON shape with `active` + `profiles`). Never call out to launch-course CLI from inside the skill for read operations — read the file. Use `launch-course` (Bash) only for `switch` (it owns the MCP rewrite invariant).
+
+## Per-subcommand behavior
+
+### status (default)
+
+Read the cred file. Print one line:
+
+```
+active: <name> (<orgSlug>/<projectSlug> @ <serverUrl>)
+```
+
+If the file is in legacy flat shape (single profile at the top), say so explicitly: `"active: (legacy single-profile shape — no name; first multi-profile op auto-migrates)"`.
+
+### list
+
+```
+courses (3):
+    name        serverUrl                              org/project
+  ★ prod        https://launchsecure-v2.vercel.app     acme/launchsecure-v2
+    staging     https://staging.launchsecure-v2.app    acme/launchsecure-v2
+    local       http://localhost:3000                  acme/launchsecure-v2
+```
+
+Sort: active first, then alphabetical. Column-align with monospace.
+
+### switch
+
+1. Sanity-check `<name>` exists in `profiles`. If not, list available names and stop.
+2. Print the BEFORE → AFTER one-liner:
+   ```
+   switching: prod (acme/launchsecure-v2 @ launchsecure-v2.vercel.app)
+             → local (acme/launchsecure-v2 @ localhost:3000)
+   ```
+3. Ask: `"Continue? (yes / no)"` — wait for `yes`. Per project memory, destructive/sticky changes require explicit confirmation.
+4. On `yes`, run `npx launch-course set <name>` via Bash. This rewrites both `.launch-secure.cred.config` (active pointer) AND `.mcp.json` (the launch-secure URL). Surface the CLI output verbatim.
+5. After switch, remind: `"Restart Claude Code to pick up the new MCP URL (or use \`/mcp reload\` if available)."`
+
+### diff
+
+Resolve `--a` and `--b` from `profiles`. Print a compact diff:
+
+```
+local         →  prod
+serverUrl     http://localhost:3000              →  https://launchsecure-v2.vercel.app
+org           acme                               →  acme
+project       launchsecure-v2                    →  launchsecure-v2
+pat           sha256:abc1…ef                     →  sha256:def0…12
+```
+
+PAT values are hashed (sha256 first 8 of each, last 2 — purely for the user to confirm they differ; never print full PATs). Same row for unchanged values.
+
+## Constraints
+
+- **Read-only by default.** Only `switch` writes; status/list/diff never mutate.
+- **Switch requires confirmation.** Per project memory, never destructive without `yes`.
+- **No PAT exfiltration.** Never print the full PAT value; show short hashes for diff identification only.
+- **Don't bypass launch-course CLI.** The `switch` MCP-rewrite logic is its responsibility — call the CLI; don't reimplement the .mcp.json mutation inside the skill (drift risk).
+- **Surface the active course on every run.** Even on `list`/`diff`, the first or last line shows which course is active so the user is never confused about session scope.

package/scaffolds/ls-marketplace/plugins/kit/skills/debug.md

@@ -0,0 +1,92 @@
+---
+description: Investigate a bug, error message, or unexpected behavior using launch-chart first (structural + variable queries), falling back to grep/Read only when the chart can't answer. When the chart gap is the reason for the fallback, files a feedback comment to LaunchSecure so the gap is monitored and fixed.
+---
+
+# /kit:debug
+
+A guided investigation workflow that defaults to chart-MCP queries and only reaches for grep/Read when the chart can't answer — and when that happens, files a one-line feedback comment in LS so the gap doesn't go unnoticed.
+
+Parse `$ARGUMENTS`:
+- **symptom** (required) — the error message, broken behavior, or question to investigate. Quoted is fine: `/kit:debug "tags don't render on the work-item drawer"`.
+- **--start=<node>** — start the investigation from a specific file/function/table id (skip the search step).
+- **--layer=<id>** — scope chart queries to one layer (ui / api / db / static).
+- **--no-grep-fallback** — refuse the grep fallback entirely (strict MCP-only).
+
+Examples:
+- `/kit:debug tags don't render on the work-item drawer`
+- `/kit:debug "TypeError: Cannot read property 'status' of undefined" --layer=ui`
+- `/kit:debug --start=src/server/comms/build-feed.ts`
+
+## Investigation loop
+
+Run this loop until the user has enough to act, OR until a fallback was filed (then surface the fallback):
+
+### Step 1 — locate the entry point
+
+If `--start` was given, jump to step 2. Otherwise, use chart to find candidate nodes:
+
+1. Extract key terms from the symptom (component name, identifier, error class). Skip noise words.
+2. Call `mcp__launch-chart__read_graph(search: <term>, type: <best-guess-type>, layer: <layer?>)`. If `type` is uncertain, omit it and filter the response by relevance.
+3. If `read_graph` returns 0 candidates, broaden: try without `type`, then without `layer`. Surface the top 3 candidates to the user as ranked options; if the user can pick, jump to step 2 with their pick.
+
+**Chart-gap detection (entry-point search):**
+- If `read_graph` returns 0 candidates AND a quick `grep` confirms the term exists somewhere in source → file a chart-gap report per the protocol below, then proceed via the grep fallback for this query only.
+- If `detect_project_stack` lists a layer (e.g. `api`) but `read_graph(layer:"api")` returns 0 nodes → file a chart-gap report (parser misconfig).
+
+### Step 2 — understand the node
+
+Once the entry point is known, query its structure and behavior:
+
+- **Structural** — `mcp__launch-chart__read_graph(node_id: <id>, hops: 1, include_edges: true)` for what it renders/imports and what depends on it.
+- **Variable / state** — `mcp__launch-chart__inspect_node(node_id: <id>, fields: ["stateVars"])` or `(filter: <key-term>)` for what state it holds, what conditions exist. Per project CLAUDE.md, ALWAYS start with `fields` or `filter` to keep the response small.
+
+Compose a hypothesis from these two reads. Example: "WorkItemDrawer fetches tags via `useWorkItemTags`, which calls `GET /api/work-items/:id/tags` — let's see what that returns."
+
+### Step 3 — follow the trail
+
+Walk the dependency edges toward the suspected cause (API endpoint, DB table, util fn, etc.). At each hop, use `read_graph(node_id, hops:1)` to find the next link. When you hit a leaf or run out of edges, narrate the chain back to the user with file:line refs from the node ids.
+
+### Step 4 — propose, don't fix
+
+End with a one-paragraph diagnosis + 2-3 candidate causes ranked by likelihood. Do NOT propose an edit unless the user asks for one — debug is investigation, not implementation. Suggest the next move: "Want me to read the route handler and confirm?" or "Want `/kit:diagram sequence`-ing the call chain?"
+
+## Grep fallback
+
+Only enter this path when:
+- the chart returned no results AND a chart-gap report was filed, OR
+- the symptom is about a non-TS asset (CSS string, env var, JSON value, comment) — the chart doesn't model these and the gap is expected.
+
+Then use `grep`/`Read` directly. Always tell the user explicitly: `"Chart returned nothing — falling back to grep (filed gap report id: …)"` or `"Symptom is in CSS/JSON, which chart doesn't model — using grep directly (no gap report needed)."`
+
+## Chart-gap protocol (shared)
+
+When the chart MCP returns less than the symptom suggests it should, file ONE feedback comment in the project's Comm Hub via `mcp__launch-secure__communication_write` (cloud LS by default):
+
+```
+resource_type: "feedback"
+title: "launch-chart gap: <one-line>"
+body: |
+  Skill: /kit:debug
+  Symptom: <one-line symptom>
+  Tool:    read_graph(search="…", layer="…", type="…")
+  Returned: 0 nodes
+  Cross-check: grep "<term>" matched <N> files (sample: <path1>, <path2>)
+  Project stack: <detect_project_stack one-liner>
+fields:
+  severity: "low"
+  source: "kit-skill"
+  skill: "debug"
+  tool: "read_graph"
+  args: { search, layer, type }
+  response_summary: { nodes_count: 0, grep_hit_count }
+```
+
+Then continue with the grep fallback for THIS query. Don't file again in the same invocation.
+
+## Constraints
+
+- **Chart-first.** Default to MCP queries; grep only when the chart cannot answer.
+- **One gap report per invocation.** Don't spam comments.
+- **No implementation.** This skill diagnoses; it does not edit. If the user wants the fix, they invoke `/fix:*` or hand the diagnosis back to you in chat.
+- **Cite by node id.** Every claim links to a chart node id (file:line via `inspect_node` lines) so the user can verify.
+- **Surface the fallback verbatim.** When grep is used, the message that grep was used (and why) must appear in the output — silent fallback hides the gap.