# Automate RunComfy Model Jobs

Running remote media models from scripts requires careful authentication, inputs, polling, and downloads. This skill provides practical RunComfy CLI workflows for those tasks.

## Install

```bash
npx skillstore add agentspace-so/runcomfy-cli
```

## Metadata

- Status: approved
- Slug: agentspace-so-runcomfy-cli
- Skillstore revision: r2
- Version status: missing
- Tree hash: 781a102c503d4a36306eb5a57bdbebae67ad94e91d1d86420deed7cec7e1141e
- Author: agentspace-so
- GitHub username: agentspace-so
- License: MIT
- Repository: https://github.com/agentspace-so/runcomfy-agent-skills/tree/main/runcomfy-cli/
- Ref: 181fdefcafd96b041926e61c4b2e306ca7e7820e
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: confirmation\_required
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands, network, filesystem
- Quality score: 38
- Quality tier: warning
- Public page: https://skillstore.pages.dev/skills/agentspace-so-runcomfy-cli
- Manifest: https://skillstore.pages.dev/api/skills/agentspace-so-runcomfy-cli/manifest

## Capabilities

- Explains npm installation and interactive or token-based authentication.
- Shows how to invoke RunComfy model endpoints with JSON or YAML inputs.
- Supports synchronous runs, deferred polling, request status checks, and cancellation behavior.
- Controls output directories, automatic downloads, and no-download responses.
- Provides machine-readable output patterns for pipelines and batch jobs.
- Documents exit codes and bounded retry handling for transient failures.

## Use Cases

- Create Media from a Terminal: Submit one image or video request, monitor completion, and save the result in a chosen directory.
- Automate Application Assets: Run repeatable model jobs from development scripts and consume machine-readable results.
- Process Prompt Batches: Submit multiple prompts with unique output folders and bounded retries for temporary failures.

## Prompt Templates

### Install and Sign In

```
Install the RunComfy CLI with npm, help me authenticate interactively, verify my identity, and stop before sending any model request.
```

### Run One Model

```
Run [model ID] with [input], save downloaded results in [directory], and explain the request status.
```

### Submit and Poll Later

```
Submit [model ID] without waiting, return its request identifier, then show how to check status and download results later.
```

### Build a Resilient Batch

```
Build a batch workflow for [prompt file] using safe JSON construction, unique output directories, machine-readable output, and bounded retries for transient errors.
```

## Limitations

- Requires a RunComfy account, valid credentials, the RunComfy CLI, and network access.
- Does not provide every model schema; users must inspect the selected model API definition.
- Remote jobs may incur charges, queue delays, timeouts, or provider failures.
- User prompts and referenced media are processed by an external service.

## Best Practices

- Inspect the selected model schema before building inputs.
- Pass user-controlled values through an input file or structured JSON builder.
- Protect tokens, use owner-only storage, and avoid logs or repository files.

## Anti Patterns

- Do not pipe an unreviewed remote installer into a shell.
- Do not ingest unrelated URLs or enable web search without explicit user need.
- Do not retry permanent errors or create unbounded retry loops.

## Security Audit

- Audited at: 2026-07-23T09:17:03.338\+00:00
- Summary: All 118 static findings are false positives caused by Markdown syntax, safe shell examples, expected URLs, a placeholder ellipsis, standard token storage, or account inspection. A separate high-severity issue remains because the skill overstates shell-injection safety when agents may interpolate untrusted prompt text. Publication should wait until all user inputs use a file or structured JSON builder.

## Stats

- Views: 55
- Downloads: 11
- Favorites: 0
- Popularity score: 0