Arc Free Worker Dispatch 1.1.0

Route tasks to free AI models via OpenRouter to save money. Use when the agent needs to delegate content writing, research, code generation, or other tasks t...

Security Scan

The skill's code, runtime instructions, and required environment variables are consistent with its stated purpose of routing tasks to free models on OpenRouter.

Purpose & Capability

Name/description match the actual behavior: the included Python script calls OpenRouter (openrouter.ai) and routes prompts to a small set of free models. Required binary (python3) and the single env var (OPENROUTER_API_KEY) are appropriate and expected.

Instruction Scope

SKILL.md simply instructs running the bundled script with task/model parameters. The script only reads the provided prompt, optional task file (tasks.json), and OPENROUTER_API_KEY; it does not attempt to harvest unrelated config files or environment variables. Caveats: batch mode will read whichever JSON file path the user supplies (no path restriction), and the code prints an error message suggesting 'credentials.txt' as an alternate location for the API key (that file is not used by the script). Output writes are validated to prevent overwriting common sensitive dotfiles, but reading arbitrary files is possible only if the user explicitly provides their path.

Install Mechanism

No installer or external downloads are used; this is an instruction-only skill with a bundled Python script. That minimizes install-time risk (nothing is fetched/installed from external URLs).

Credentials

Only OPENROUTER_API_KEY is required, which is proportionate to calling the OpenRouter API. No unrelated credentials, config paths, or excessive env variables are requested.

Persistence & Privilege

Skill is not always-enabled; it does not request persistent system privileges, does not modify other skills or global agent configuration, and uses standard, user-invoked behavior.

Assessment

This skill appears coherent and implements exactly what it claims: a Python script that sends prompts to OpenRouter using the OPENROUTER_API_KEY. Before installing/use: 1) Treat your OPENROUTER_API_KEY like any secret — the skill will send any prompt content (including file contents you pass) to openrouter.ai, so do not include passwords, private keys, or other secrets in prompts or task files. 2) The script prevents writing to common sensitive dotfiles, but batch mode will read any JSON file path you provide — avoid pointing it at sensitive system files. 3) The author/source is unknown and there's no homepage; if you need a higher assurance, review the included scripts yourself or run them in a restricted environment. 4) Note the script sets an HTTP referer header referencing arcself.com (cosmetic but observable by the remote API). If you accept those caveats, the requested permissions and behavior are proportionate to its purpose.

Free Worker Dispatch

Delegate tasks to free AI models via OpenRouter. Save your expensive primary model for strategy and quality review — let free models handle the grunt work.

Why This Exists

Running everything through Claude Opus or GPT-4 costs real money. Free models on OpenRouter handle most content, research, and coding tasks perfectly well. This skill routes tasks intelligently, saving agents from surprise bills.

Available Free Models

Model	Best For	Context
stepfun/step-3.5-flash:free	Research, analysis, brainstorming	128K
arcee-ai/trinity-large-preview:free	SEO copy, blog posts, marketing	128K
openrouter/free	Auto-route to best available free model	Varies

Commands

Dispatch a task to a free model

python3 {baseDir}/scripts/dispatch.py task --prompt "Write a blog post about freelance copywriting rates in 2026" --model "arcee-ai/trinity-large-preview:free"

Dispatch with auto-model selection

python3 {baseDir}/scripts/dispatch.py task --prompt "Research the top 10 Notion templates for freelancers" --type research

List available free models

python3 {baseDir}/scripts/dispatch.py models

Check model status (is it up?)

python3 {baseDir}/scripts/dispatch.py status --model "stepfun/step-3.5-flash:free"

Dispatch with output to file

python3 {baseDir}/scripts/dispatch.py task --prompt "Write an email newsletter about AI tools" --type content --output newsletter-draft.md

Batch dispatch (multiple tasks)

python3 {baseDir}/scripts/dispatch.py batch --file tasks.json

The tasks.json format:

[
  {"prompt": "Write a product description", "type": "content"},
  {"prompt": "Research competitor pricing", "type": "research"},
  {"prompt": "Generate a Python script for...", "type": "code"}
]

Task Types

The --type flag auto-selects the best free model:

Type	Model	Why
research	stepfun/step-3.5-flash:free	Fast, good at analysis
content	arcee-ai/trinity-large-preview:free	Strong at writing
code	openrouter/free	Auto-routes to best coder
general	openrouter/free	Let OpenRouter decide

Output

Results are printed to stdout by default. Use --output <file> to save to a file. Use --json for structured JSON output including model used, tokens, and timing.

Tips

• Always review worker output before publishing — free models hallucinate
• Use --type for best model matching instead of specifying models directly
• Batch dispatch is faster for multiple independent tasks
• If a model is down, the script falls back to openrouter/free