Skip to content

meshy-dev/meshy-3d-agent / skills/meshy-3d-printing

meshy-3d-printing

Verified

Generates and prepares models for physical 3D printing with the Meshy CLI: white and multicolor prints (3MF, STL, OBJ), printability analysis and repair, splitting into parts, Creative Lab products such as figures, lamps and keychains, sizing, and slicer handoff (Bambu Studio, OrcaSlicer, PrusaSlicer, Cura). Use when the user wants something 3D-printed or asks for print or slicer files.

FileKindSize
SKILL.mdskill md6.7 KB
references/creative-lab.mdmarkdown7.8 KB
references/delivery.mdmarkdown9.2 KB
references/printing.mdmarkdown16.7 KB
references/setup.mdmarkdown13.9 KB
references/troubleshooting.mdmarkdown9.6 KB
references/visual-hosts.mdmarkdown1.7 KB

references/troubleshooting.md

View on GitHub
# Recover without duplicating a task

Every command uses `--format json --no-update-check` and prints one envelope. Read the process
exit code and the envelope's `error`, `result` and `warnings`; preserve any task ID, operation
ID, downloaded files and recovery command even when `ok` is false. Auth results are in
`result` as well; see [setup.md](setup.md).

Contents: runner problems · authentication · exit codes · unknown submission, failure and
timeout · deleting a task · local files.

## Runner problems

- **`meshy: command not found`** or **a version older than 0.8.6**: install the newest 0.8.x as
  Setup describes. If the install is refused, use the temporary package
  (`npm exec --yes --package=meshy-cli@0.8.6 -- meshy …`) and carry on with the same job.
- **A global 0.9 or newer** answers `--version`. Do not uninstall or downgrade it; use the
  temporary package for this work and mention the mismatch once.
- **Node older than 22.12** fails the CLI's own runtime gate. Report the required version and the
  installed one; do not try to relax it or run an older CLI.
- **`npm exec` cannot reach the registry**: report the network/proxy error. A cached package
  still runs offline; an uncached one cannot be conjured.

## Authentication and connectivity

`auth status` exit 3 is ambiguous: `authenticated: false` means no usable credential was
resolved, while `authenticated: true, verified: false` also covers network and server failures.
For the latter, use the hint or a free `meshy balance --format json --no-update-check` to
obtain a typed error. Keep the same explicit `--api-key-file` if that is the selected source;
status does not honor it.

- **Network failure during verification**: say it is a connectivity problem, keep the stored
  credential, and retry the check when the network is back. Do not start a new authorization.
- **Auth rejection**: use the login flow in [setup.md](setup.md). Do not clear all profiles.
- **Refresh failure**: a still-valid access token may keep working; an expired one cannot.
  Check connection/server errors before asking the user to authorize again.
- **Login code expired or denied**: rerun the same `auth login --device` for a fresh code, and
  show the new URL and code. Nothing paid is retried because a login timed out.
- **Works in another terminal**: compare status `source`, active profile, OS user and
  config/API environment. An environment key overrides a stored OAuth session.
- **`device_flow_not_supported`**: this API host has no device endpoint — the user logs in
  from their own desktop terminal, or supplies their own key privately.

## CLI 0.8 exit codes

| Exit | Meaning | Recovery |
|---|---|---|
| 1 | Generic/server/protocol error or failed task | Inspect the error and any known task before deciding next steps. |
| 2 | Usage or operation conflict | Correct arguments/account/payload mismatch; do not invent a fresh operation ID to bypass a conflict. A local STL/OBJ/FBX `--model-url` lands here: pass it as an http(s) URL instead. |
| 3 | Authentication | Use the source-aware checks above; status also uses this for failed network verification. |
| 4 | Validation | Correct the rejected input or explain the unsupported request. |
| 5 | Not found | Check the resource, task ID and account; a task older than Meshy's 3-day retention is gone (see Delivery). |
| 6 | Rate limit | Back off for read/status requests; respect server guidance. A create refused with a message starting `NoMorePendingTasks` made no task: too many of the account's tasks are queued or running, so wait for one to finish, then submit once. |
| 7 | Network | Retry safe reads after connectivity recovers; inspect submission state before any create. |
| 8 | Timeout | Resume waiting on an existing task; a login timeout needs a new authorization flow. |
| 9 | The account cannot pay for this task | Too few credits, a Free plan (creating tasks needs a paid plan), a per-key monthly credit limit, or a model that needs a Pro plan or above (`nano-banana-pro`, or `--pose-mode` on an older nano-banana model). Read the message, report the actual cause, and do not resubmit the same request. `error.details.body.code` `plan_upgrade_required` (HTTP 402) is the plan, not the balance, even when the hint says out of credits and `error.recovery` suggests a top-up: credits do not lift it. If the refused task used one of those models, retry once on `gpt-image-2-5-flare`. Otherwise, or if that retry gets the same code, tell the user that creating tasks needs a Pro plan or above (plans and prices: <https://www.meshy.ai/pricing>), and do not suggest a top-up or an upgrade. |
| 10 | Submission outcome unknown | Reconcile first; never automatically resubmit. |
| 11 | Local I/O | Fix the output path/permissions, or choose another path when the target holds files that are not that task's own; then rerun the same `-o` or `download`. |
| 12 / 13 | Check failed / unknown | Inspect check details; do not present an unknown result as a pass. |
| 14 | Conflict: the task is running | A delete of an `IN_PROGRESS` task; see *Deleting a task* below. |
| 130 | Interrupted (Ctrl-C, SIGTERM or SIGHUP) | Preserve task/operation context and reconcile before continuing. |

These are 0.8 codes, not HTTP status codes. Some OAuth errors use generic exit 1; inspect
their message and hint, and `error.details.oauth_error` when it is present, as well.

## Unknown submission, failure and timeout

A create may have reached Meshy even if the response was lost. On `submission_unknown`
(exit 10), retain the original operation ID (`result.submission.operation_id`, the local journal
record of that create) and follow `error.recovery.command` to inspect the resource list.
Correlate candidate tasks with the submitted request, then close the record:

```bash
meshy operations list --state unknown --format json --no-update-check
meshy operations resolve OPERATION_ID --task-id TASK_ID --format json --no-update-check
meshy operations resolve OPERATION_ID --not-created --format json --no-update-check
```

`--task-id` binds the task the list showed: a replay of that operation ID returns it with no new
request, and you carry on with its `get`/`wait`. `--not-created` releases the ID when the list
shows no matching task; the original create run again with `--operation-id OPERATION_ID` then
submits it, which is a new paid create, so only within the user's existing authorization. A record still at `started` (the process was killed
mid-request) is resolved the same way, and `operations show OPERATION_ID` reads one record. These
commands are local and need no sign-in. If the outcome stays ambiguous, report that uncertainty
and wait for a decision. Do not generate another operation ID or resubmit automatically.
Changing authentication or account during recovery can also cause an operation conflict.

If a task ID is already known, retrieve or wait on **that same task** with its resource's
`get`/`wait`. A polling timeout, or a task paused at 99%, does not cancel the remote task.
Resume waiting; do not create a replacement. `wait` already rides out a rate limit, a server
error or a slow response on a status check and polls again, so one of those is not a reason to
stop either.

A terminal `FAILED` task is different from a failed poll. A `get` of one is a successful query
(exit 0, `ok: true`): read `result.task.status` and `result.task.task_error`. A `wait` that ends
`FAILED` is `ok: false` with exit 1 and the whole task kept in `result`. Explain the server's
reason. A replacement generation is another potentially billable operation: retry only within the user's
existing retry/budget authorization, otherwise present that concrete choice. Do not apply
generic HTTP retry loops to creates, repairs, conversions or whole multi-step pipelines.

## Deleting a task

Delete only what the user asked to remove. A `PENDING` task is deleted and refunded. A task
that is already running cannot be stopped: delete exits 14 (`conflict`), the task keeps
running and its credits are already spent. Wait for its result instead; delete it afterwards
only if the user still wants it gone, which is permanent and refunds nothing.

Creative Lab tasks are not reached by `meshy delete`; use `meshy creative-lab PRODUCT STAGE
delete` (`meshy creative-lab fidget-collapsible delete` for the single-stage product), and it is
a cancel, not a delete. `PENDING` is refunded. A running task is stopped, losing both its result
and its credits. A finished one cannot be cancelled (exit 4) or deleted at all. The task stays
listed as `CANCELED` (`canceled: true, deleted: false`).

## Local files

A write refused as outside the workspace means the path escaped `WORKSPACE` (or a symlink pointed
out of it). Fix the path or widen the workspace deliberately — do not drop `--workspace`. An
existing destination file is refused rather than overwritten (exit 11, before anything is
downloaded): choose a new name, or for `download` pass `--overwrite` when the user's instruction
covers replacing that file. `-o` has no overwrite flag; it replaces only the files of the same
task's earlier `-o`, which its sidecar names.

If generation succeeded but a download or a local step failed, keep the successful task and
repair only the failed step. A part-way `-o` transfer (exit 7 or 130) still writes its sidecar for
the files that landed, and `error.recovery.command` is the same `get … -o …` with absolute paths:
run it as printed once the cause is gone, and it replaces what landed and fetches the rest. A
process killed hard mid-transfer leaves files without a sidecar; the rerun then refuses them, so
choose a new path, or ask the user before deleting those partial files. A `download` with the same selection can likewise be
re-run for the rest. Never restart generation to obtain a missing file.
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall