Skills

Skill 23 of 34

Comfyui Workflows

Inspect and run ComfyUI workflows on RunComfy GPU deployments. Use for workflow inputs, deployment selection, serverless inference, request status and output retrieval.

2 minutes · 413 words · 4 sections

Install

$ npx skills add runcomfy-com/skills --skill comfyui-workflows

$ npx skills add runcomfy-com/skills

» /plugin marketplace add runcomfy-com/skills

The first command installs just this skill, by the name in its SKILL.md; the second installs the whole repository.

Run ComfyUI workflows on RunComfy

Use RunComfy’s connected MCP tools. Authenticate through the client’s sign-in control when required; never ask for credentials in chat. General questions about local ComfyUI do not require account access or a cloud run.

Inspect before executing

Call list_deployments and select the deployment the user owns and intends to use. Use get_deployment with include_payload: true for its actual workflow graph, inputs, defaults, GPU configuration and current state; the default response omits the full payload.
Map the user’s requested changes to the workflow’s real input names and node IDs. Do not guess which node controls the prompt, seed, dimensions, steps or file input. Preserve unspecified values.
File handling depends on the deployment schema and submit_request contract. Follow the exposed input contract rather than assuming every ComfyUI file field accepts an arbitrary URL. Do not publish private inputs without permission.
Use get_balance and current deployment pricing/configuration to explain the available cost estimate. GPU startup, execution and keep-warm time may affect charges; a job timestamp alone does not establish billed runtime.
Present deployment ID, requested overrides, expected outputs, GPU settings and budget/stop condition. Reuse the user’s existing authorization. Creating, updating or deleting a deployment changes persistent resources; do that only when requested or necessary within the agreed task.

Execute and recover

Submit exactly once with submit_request, using the actual deployment ID and schema-valid overrides. Prefer asynchronous completion when the tool exposes it.
Save the request ID. Poll get_request_status and retrieve get_request_result for that ID. These tools are for deployment requests; hosted Model API jobs use the separate get_model_request_* tools.
A timeout is not evidence that the run was rejected. Reconcile the existing request before submitting any replacement.
Use cancel_request for a user cancellation or an agreed stopping condition. Explain that cancellation may not undo already incurred charges.
Return actual output URLs and request ID. Check the deployment’s state after completion when the task includes releasing GPU resources or controlling ongoing cost. Do not delete the deployment as routine cleanup.
Describe standby and keep-warm state accurately. Report actual wallet movement only after a relevant recheck, and do not attribute concurrent account activity to this one job.
Use call_instance_proxy only for a specific authorized action whose route and effect are understood; it is not a read-only debugging shortcut.

Useful requests

“List my RunComfy deployments and inspect the image workflow’s inputs. Do not run or change anything.”
“Run one image on this deployment with these overrides, within the budget I approved, and return its output.”

ComfyUI workflows (opens in a new tab) · Serverless API documentation (opens in a new tab)

About this skill

Trigger

Inspect and run ComfyUI workflows on RunComfy GPU deployments. Use for workflow inputs, deployment selection, serverless inference, request status and output retrieval. Review the actual workflow schema and GPU settings before paid execution or deployment changes.

The verbatim description from this skill’s front matter — the string an agent matches on to decide whether to load it.

Colophon for this issue

Colophon

How this issue was made

Source

github.com/runcomfy-com/skills, branch main, last pushed 15 September 2026.

Licence

MIT — the text of every skill is reproduced unmodified, frontmatter included, under the upstream licence.

Discovery

34 skills found by walking the repository tree for SKILL.md, not by matching a directory convention. 2 distinct layouts observed: */SKILL.md, plugins/runcomfy/skills/*/SKILL.md.

Issue colours

Resolved from a deterministic hash of the owner name. Two accent tones are generated per issue and each is proven against its own ground before it ships: a single accent that passes AA on both light and dark paper is arithmetically impossible.

Heading repairs

1 repair applied to this skill so the document has one h1 and no skipped levels:

Shifted “4 headings” from h1 to h2 so the skill title is the only h1.

Spec compliance

2 editorial notes across 2 of 34 skills. They are printed in the margin of each skill rather than as errors here.

AI Music — 1 note
Runcomfy CLI — 1 note

Images

Images inside a skill come from the upstream repository. Where the author gave no alternative text we mark the image decorative rather than inventing a description — a plausible caption we made up is worse than none for the reader who depends on it.

Marketplace

A plugin manifest is published at .claude-plugin/marketplace.json by RunComfy, declaring 1 plugin. It is read for editorial metadata only — never as the skill index, which is always the repository tree.

Signal

This repository is unranked on skills.sh, which says nothing about it: the leaderboard covers a fraction of published skills.

Agent surfaces

The whole issue is available as one markdown document at /runcomfy-com/skills.md, and each skill at its own .md URL.

Publication

Set by Skills Docs from the source repository. Body text is Literata at the reader’s chosen size and measure; code is Geist Mono. Nothing on this page was written by us except this paragraph.

Run ComfyUI workflows on RunComfy

Call list_deployments and select the deployment the user owns and intends to use. Use get_deployment with include_payload: true for its actual workflow graph, inputs, defaults, GPU configuration and current state; the default response omits the full payload.

Map the user’s requested changes to the workflow’s real input names and node IDs. Do not guess which node controls the prompt, seed, dimensions, steps or file input. Preserve unspecified values.

File handling depends on the deployment schema and submit_request contract. Follow the exposed input contract rather than assuming every ComfyUI file field accepts an arbitrary URL. Do not publish private inputs without permission.

Use get_balance and current deployment pricing/configuration to explain the available cost estimate. GPU startup, execution and keep-warm time may affect charges; a job timestamp alone does not establish billed runtime.

Present deployment ID, requested overrides, expected outputs, GPU settings and budget/stop condition. Reuse the user’s existing authorization. Creating, updating or deleting a deployment changes persistent resources; do that only when requested or necessary within the agreed task.

Submit exactly once with submit_request, using the actual deployment ID and schema-valid overrides. Prefer asynchronous completion when the tool exposes it.

Save the request ID. Poll get_request_status and retrieve get_request_result for that ID. These tools are for deployment requests; hosted Model API jobs use the separate get_model_request_* tools.

A timeout is not evidence that the run was rejected. Reconcile the existing request before submitting any replacement.

Use cancel_request for a user cancellation or an agreed stopping condition. Explain that cancellation may not undo already incurred charges.

Return actual output URLs and request ID. Check the deployment’s state after completion when the task includes releasing GPU resources or controlling ongoing cost. Do not delete the deployment as routine cleanup.

Describe standby and keep-warm state accurately. Report actual wallet movement only after a relevant recheck, and do not attribute concurrent account activity to this one job.

Use call_instance_proxy only for a specific authorized action whose route and effect are understood; it is not a read-only debugging shortcut.

“List my RunComfy deployments and inspect the image workflow’s inputs. Do not run or change anything.”

“Run one image on this deployment with these overrides, within the budget I approved, and return its output.”