Skip to content

Razshy/Wiggle / mnt-skills/examples/pages

ALWAYS start here when the user wants a page OR a doc: page, living page, doc, document, living doc, bongo or bongo doc (the old name; new ones are made here, not by a Bongo connector), shared or collaborative doc, write-up, one-pager, memo, brief, proposal, plan, spec, report, meeting notes, notes, checklist, tracker, or anything to read, edit, comment on and share together — 'write this up', 'put this in a doc', 'draft a brief', 'jot this down somewhere we can share and edit', 'turn this file into a page' — the DEFAULT for any of these when no file format is named; a Word/.docx, PDF, .pptx or .xlsx FILE asked for by name goes to that format's skill instead. Also whenever the pages connector is about to be used to create a page. Not for READMEs, code or docstrings. A lowercase /pages as its own word anywhere in the message (not in a URL, path or longer name) invokes this skill whatever else is asked — a summary, an answer, a comparison, a plan: the page is the reply.

FileKindSize
SKILL.mdskill md11.1 KB
charts-and-data.mdmarkdown11.0 KB
failure-branches.mdmarkdown9.7 KB

failure-branches.md

View on GitHub
# When a step does not go as described — decide by what the tool answered

Companion to SKILL.md, whose steps 2, 3 and 5 already carry the FIRST MOVE for each case below; this file is the procedure and the exact words. Nothing here makes a substitute: no HTML file, no hand-published artifact, no .md/.docx presented as the deliverable, never the same refused call re-sent unchanged. Only CREATING ever stops; an existing page is still read, revised and commented on through the connector (SKILL.md, Afterwards), also from a session that cannot create.

## Step 2, road (B) — the artifact-tool call (decide from its parameter list and its answer; never probe it with the type's link in `file_path` or `url`)

- **`type_url` refused as an unexpected parameter, or answered "not found", "not available", "not enabled" / "denied"; or no URL came back** → nothing was created. Fall to road (C): single-write (`create:{"name"}`, no `artifact`) and the link once in the hand-over. Never publish an HTML file, a placeholder page or any other artifact as a substitute, never hand over the type's link.
- **The answer says the artifact "may or may not have been created"** → the tool's `list` action first, before anything else: if the list shows an artifact with step 1's title, that IS the artifact — use its URL exactly as listed as step 3's `artifact`; if it shows none, make the `type_url` call once more. Never a second artifact for the same page, never a guess at the URL.
- **"…this session targets staging"** → the same call with SKILL.md's staging `type_url`, once; nothing was created yet.
- **"`session_context` is required"** → the same call plus a one-line `session_context` (what the page is for); still no `file_path`, `files` or `capabilities`. Nothing was created yet.
- **No `list_types` (or it errors)** → expected: never enumerate types; the `type_url` in SKILL.md is the whole input.

## Step 3 — the birth batch (the one pages-connector `batch` call)

- **No pages connector at all (no `batch`/`create`/`update` tools of it), or the only pages-like tools are named `v3_…`** → this session cannot create a page (the `v3_…` connector is an older one that serves existing pages only) — not a token, auth or transient fault, whatever `whoami` says. Make NO page, no substitute file. Say: "This session doesn't have the pages connector, so I can't create your page here — connect it (Settings → Connectors) or run /pages where it is connected, and it will build the page; I can draft the outline here in chat if useful."
- **Roads (A)/(C): `batch` refused over `container.create` naming an `artifact` key** → you added one: on these roads the birth is `container: {"kind":"project","create":{"name":"<title>"}}` and nothing else in `create`. Send it once, correctly.
- **Road (B): the `artifact` VALUE is refused ("expected the artifact's URL … or its id", "you cannot bind a page to this artifact")** → pass exactly what the Artifact tool returned (the full URL, never the type's link, never a parsed or invented id) and send the batch once more. "cannot bind … not yours" after a `list`-found artifact → the `type_url` call again for a fresh artifact, bind that one. `batch` DENIED over the `artifact` KEY itself (`artifact-attest`, "artifact: unknown argument") → this connector cannot bind: send the birth WITHOUT `artifact` (road C) and give the link once; the road-B artifact then stays empty — do not hand over ITS link.
- **A refusal saying an argument "expected an object/array — got a STRING"** → you passed `container` or `batch` as text: send them as real JSON values (an object and an array), same content, once.
- **A member is refused for its shape** (`intent-shape` / `value-shape` naming the file member or the pointer member) → send the three members exactly: the tab `{"$lid":"f","$tab":{"name":"<title>"},"verb":"create","object":"file","engine":"json","payload":{"engine":"json","value":{"birthName":"Doc.ldoc"}}}`, then SKILL.md's prose node member, then the pointer `{"verb":"update","ref":{"object":"file","id":"$lid:f"},"engine":"json","payload":{"engine":"json","kind":"value","value":{"birthName":"Doc.ldoc","content":{"kind":"node","id":"$lid:n"}}}}` — with the same `container`.
- **`name` refused (generic or empty title)** → step 1 again: a title its reader would say, then the same batch once.
- **The call errored or timed out with no acknowledgement ("may or may not have been created")** → `list_projects` with `query:"<title>"` first: if the page is there, it IS step 3's page — `read` it (ref `{object:"project", id}`) for its tab and prose node and carry on at step 3/4 with that id; if it is not, send the birth once more. Never a second page for the same ask, never a guessed id.
- **`created` answers `verdict:"deny"`, `reason:"exists"` with `minted` + `bound: true`** → that page is already YOURS (provisioned before your turn, e.g. from the + button): never send the birth again; address it with `container:{"kind":"project","id":"<minted>"}`. If `created` also carries `lids`, the page already has its tab — write step 3's document into THAT tab with an `update` (step 4's call shape) on `lids.n` (the tab's prose root; `lids.f` is the tab); no `lids` ⇒ `read` the container first to find its tab and that tab's prose root, then the same `update`. Then steps 4–5 as usual (`bound: true` there counts as step 3's). A bare `exists` (no `minted`, no `bound`) on road (B) ⇒ the id is held by someone else: the `type_url` call again for a fresh artifact, bind that one.

## Step 3, after the ack — opening the page (by road)

- **`created.bound: true`, trailer says the page is already open beside the user (road B, or a host that opened it itself)** → do nothing to open it; no link in the hand-over unless they ask or say they cannot see it (then give the viewer URL once).
- **`created.bound: true`, trailer says the viewer is at a URL and NOT open beside the user yet** → road (A): your Artifact tool's `open` action with that URL (no `file_path`, `files`, `type_url`, `capabilities` — you are opening, not publishing); road (C), or `open` refused: the link once as the hand-over's first line. Never publish anything to "make" the viewer: it already exists.
- **`open` answered an error** → fall back to giving the link once; do not retry `open`, do not publish.
- **`created.bound: false`, or a `notice` in the acknowledgement** → the content is saved but the page is NOT openable from any link yet: never say "ready", "open", "live" or "done" and hand over no link as if it worked. If the notice names a next step, follow it once. Otherwise say "Your page's content is saved, but it can't be opened yet —" plus what the notice says is missing, and offer to finish it from a session that has it. Keep filling (step 4) unless the notice says not to. Never delete the page to clean up.
- **No `frame` / no URL anywhere in the ack but `bound: true`** → the link is `https://claude.ai/code/artifact/<created.minted>`; treat it as "not open yet".

## Step 4 — fills and later writes

- **`unknown_id` / `relative_id` / "bare clock"** → the id was not FULL: compose `<session>.<clock>` from the acknowledgement's `session` and the `keys.empty` clock (e.g. `mwz47r07s9v.51`); a `read` prints later ids short (`.51`) under a full id like `mwz47r07s9v.3` — send `mwz47r07s9v.51`, never `.51`. The whole fill call, for reference: `{"ref":{"object":"node","id":"<node id>"},"engine":"prose","container":{"kind":"project","id":"<page id>"},"payload":{"engine":"prose","ops":[{"op":"replace","target":{"kind":"blocks","ids":["<session>.<clock>"]},"ifHash":"e3b0c442","with":{"from":{"kind":"inline","content":"<the section's real markdown>"},"as":"markdown"}}]}}`.
- **`forbidden {why:"role"}` / "container required" on an `update` or `read`** → the call lacked `container:{"kind":"project","id":"<page id>"}` (the page id is `created.minted`, the UUID ending the viewer URL): add it, same ops.
- **`guard_mismatch`** → a person already typed into that block: their text stays. Resend the call without that op (the refusal shows their block), fold your addition beside theirs (an `insert` after it), and tell the user their words were kept.
- **`row_too_large`** → one call carried too much text: split the fill into more `update` calls (2–4 sections each), or put a very long body up by the file road and `replace` from the blob.
- **`unsupported_construct`** → the markdown used one of the three spellings that refuse a call: `$…$` math (write it as code), a footnote `[^1]` (parenthesize), or an image whose src is not `blob/<id>` (upload it by the file road, or link it). Fix and resend once.
- **`find_none` / `find_ambiguous`** on a `find` target or comment anchor → the words moved or occur twice: quote more words, or add the `parentId` / `nth` the refusal's `data.matches` offers.
- **A `read` answers `complete: false` (a `cap` gap)** → the page is longer than one reply: continue with the `data.next` it returned (same read, `from` = `data.next.from`) until no `next`, before summarizing or answering from it — never treat the first part as the whole.
- **`upload_asset` refuses the file's TYPE ("unsupported asset type …", listing what it takes)** → tell the user so plainly first, then the last bullet of the file road in charts-and-data.md (a dataset may be charted inline; a document is converted and uploaded, never pasted in unasked).

## Step 5 — the hand-over

- **The user says they cannot see the page / asks for the link** → give `[<title>](<viewer URL>)` (again, if it was already the hand-over's first line); the page id is the UUID in it.
- **No openable page exists at the end (no connector, `bound: false`, a `notice`)** → the reply names what is missing and where to run /pages; it contains no "ready", "live", "open it", "done", and no link presented as a working page.
canvas-designanthropics/skillsCreate beautiful visual art in .png and .pdf documents using design philosophy. You should use this skill when the user asks to create a poster, piece of art, design, or other static piece. Create original visual designs, never copying existing artists' work to avoid copyright violations.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
docxanthropics/skillsUse this skill whenever the user wants to create, read, edit, or manipulate Word documents (.docx files) or Word templates (.dotx files). Triggers include: any mention of 'Word doc', 'word document', '.docx', '.dotx', or requests to produce professional documents with formatting like tables of contents, headings, page numbers, or letterheads. Also use when extracting or reorganizing content from .docx or .dotx files, inserting or replacing images in documents, performing find-and-replace in Word files, working with tracked changes or comments, or converting content into a polished Word document. If the user asks for a 'report', 'memo', 'letter', 'template', or similar deliverable as a Word or .docx file, use this skill. Do NOT use for PDFs, spreadsheets, Google Docs, or general coding tasks unrelated to document generation.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
pdfanthropics/skillsUse this skill whenever the user wants to do anything with PDF files. This includes reading or extracting text/tables from PDFs, combining or merging multiple PDFs into one, splitting PDFs apart, rotating pages, adding watermarks, creating new PDFs, filling PDF forms, encrypting/decrypting PDFs, extracting images, and OCR on scanned PDFs to make them searchable. If the user mentions a .pdf file or asks to produce one, use this skill.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
pptxanthropics/skillsUse this skill any time a .pptx or .potx file is involved in any way — as input, output, or both. This includes: creating slide decks, pitch decks, or presentations; reading, parsing, or extracting text from any .pptx or .potx file (even if the extracted content will be used elsewhere, like in an email or summary); editing, modifying, or updating existing presentations; combining or splitting slide files; working with templates (.potx), layouts, speaker notes, or comments. Trigger whenever the user mentions "deck," "slides," "presentation," or references a .pptx or .potx filename, regardless of what they plan to do with the content afterward. If a .pptx or .potx file needs to be opened, created, or touched, use this skill.Stars: 198kUpdated Claude CodeCodexVerifiedInstall