qwencloud-model-selector

name: qwencloud-model-selector description: "Recommend the best Qwen model and parameters. TRIGGER when: choosing between Qwen models, comparing Qwen model pricing, understanding Qwen model capabilities, asking which models are available on Token Plan, viewing cost history, when an execution skill needs model selection advice, or user explicitly invokes this skill by name (e.g. use qwencloud-model-selector). DO NOT TRIGGER when: non-Qwen model discussions (OpenAI, Gemini, etc.), general AI questions unrelated to Qwen." compatibility: "Advisory skill. CDN snapshot lookup has no local dependency; real-time model and account queries require QwenCloud CLI and Node.js >= 18. Cursor: auto-loaded. Claude Code: read this skill's SKILL.md before first use."

Qwen Model Selector (Advisor)

This skill operates in two modes:

1. Interactive advisory — asks diagnostic questions to recommend the right model (see Diagnostic Flow).
2. Cross-skill resolution — provides a fast-path model lookup for execution skills that need a model decision without user interaction. Fetch the CDN recommendations first; if CDN access fails, use the local fallback.

Do not fabricate model names — only recommend models listed in the CDN catalogs or returned by CLI. This skill is part of qwencloud/qwencloud-ai.

🚫 CRITICAL — Never override a user-specified model. If the user (or the requesting execution skill) has already explicitly specified a model, this skill's job is done for that request: use the model as given. Do NOT recommend a "better suited" or newer model in its place, and do NOT add parameters (thinking mode, style hints) the user did not ask for. Recommendations and diagnostic flows below apply only when no model has been specified. You may verify availability and, if the model is unavailable/restricted, inform the user and suggest alternatives — but the user makes the final call.

Skill directory

Load on demand. Fetch the documented CDN model catalogs when model lists, defaults, or recommendations are needed. Do not fetch other external URLs unless the user explicitly asks for the latest data.

Every path under cdn/ is a local fallback, not the primary source. For cdn/<path>, first fetch https://alioth-intl.alicdn.com/skills-info/models/<path> and use the local file only if that request fails.

Prerequisites

QwenCloud CLI is strongly recommended — it is the authoritative real-time data source for model availability, pricing, and quotas. Verify with:

qwencloud version

If not installed:

npm install -g @qwencloud/qwencloud-cli

Node.js >= 18 required. Without CLI you can still answer point-in-time navigation and model questions from the CDN catalogs, but you cannot verify current availability, exact current prices, or account quotas.

Security & Credential Model

QwenCloud has two independent credential systems — never confuse them:

Red lines (apply to both):

• NEVER output any credential value in plaintext. Use variable references; report only status ("set" / "not set", "valid" / "invalid"). Never display .env or config file contents.
• NEVER conflate the two systems. When CLI returns Not authenticated / AUTH_REQUIRED, run the 3-step device-flow login (see cli-usage.md). DO NOT ask the user for an API key, and DO NOT try to set $DASHSCOPE_API_KEY to fix CLI auth.

Detecting Key Type

Before recommending models, detect the billing mode from the configured API key (outputs token-plan, payg, or not-set — never the key value itself):

python3 -c "
import os
from pathlib import Path
env_file = Path('.env')
if not env_file.exists():
    for parent in [Path.cwd()] + list(Path.cwd().parents):
        if (parent / '.git').exists() or (parent / 'skills').is_dir():
            env_file = parent / '.env'
            break
if env_file.exists():
    for line in env_file.read_text().splitlines():
        line = line.strip()
        if not line or line.startswith('#') or '=' not in line:
            continue
        k, v = line.split('=', 1)
        k, v = k.strip(), v.strip().strip('\"').strip(\"'\")
        if k in ('QWEN_API_KEY', 'DASHSCOPE_API_KEY') and k not in os.environ:
            os.environ[k] = v
key = os.environ.get('QWEN_API_KEY') or os.environ.get('DASHSCOPE_API_KEY') or ''
print('token-plan' if key.startswith('sk-sp-') else 'payg' if key else 'not-set')
"

Windows: If python3 is not available, use python. Alternatively, save the snippet to a temporary .py file and run it.

Data Resolution Order

Match the user's question to the right data source. Do not fall back to a lower tier without trying the recovery actions in the higher tier first.

Diagnostic Flow (Interactive Advisory)

Ask the user (in order):

1. Content type? — text / image / video / audio / vision
2. Primary task? — generation / understanding / coding / reasoning / translation
3. Priority? — quality vs speed vs cost
4. Input size? — short / medium / long context
5. Structured output? — JSON / function calling needed?

Default Recommendations

Before choosing a default or recommending a model, fetch and read the current CDN recommendations. For a Token Plan key, also read the Token Plan catalog linked above and restrict candidates to it.

• If the user explicitly specifies a model, preserve it and only report genuine availability conflicts.
• Otherwise, match the strongest task signal first, then choose the appropriate quality, speed, or cost tier.
• Preserve the selected model's documented thinking default unless the user explicitly requests an override.
• If CDN access fails, use the local recommendations.

Degradation: If this skill is not loaded, each execution skill loads its default from CDN and falls back to its bundled configuration. This protocol is purely additive and never blocks execution.

CLI Quick Reference

Auth required. All models and usage commands need an active CLI session (browser device-flow login — NOT the API key). If the command returns Not authenticated / AUTH_REQUIRED:

1. Run the 3-step device-flow login in cli-usage.md (proactively open the verification URL using the OS-appropriate command, then poll immediately).
2. Retry the original command after success.
3. DO NOT ask the user for $DASHSCOPE_API_KEY / $QWEN_API_KEY — those are for model API calls, not CLI session. See Security & Credential Model above.
4. DO NOT silently fall back to snapshots.

Display rules: Parse --format json output and present a human-readable summary; never dump raw JSON. Display --format text output as-is, then add analysis after ---. See cli-usage.md for details.

CLI Error Handling — Quick Guide

When CLI fails, classify first, recover, then retry. Never silently fall back to snapshots.

Full classification, signals, and example flows: error-handling.md.

Pricing & Cost Estimation

• Latest pricing: Run qwencloud models info <model> --format json first; use the CDN pricing reference linked by pricing.md only as fallback. Never invent a price.
• Mandatory disclaimer: Every cost-related answer must end with the disclaimer in pricing-disclaimer.md (Chinese or English version, matching the user's response language). Omitting the disclaimer is a critical failure.
• Free quota: Never assume free quota is available — use qwencloud usage free-tier to verify or direct the user to the console.
• Usage / billing queries: Direct the user to the appropriate console page — see the table in pricing-disclaimer.md.

Update Check

When the user asks to check for updates ("check for updates", "check version", "is there a new version", "update skills"):

1. Find qwencloud-update-check: Look for qwencloud-update-check/SKILL.md in sibling skill directories.
2. If found — run: python3 <qwencloud-update-check-dir>/scripts/check_update.py --print-response and report the result. Use --force if the user asks to force-check.
3. If not found — run qwencloud version --check and report the result.

Anti-Patterns

• Never override a user-specified model — if the user explicitly chose a model, recommend nothing else for that request; only verify availability. Suggestions are allowed only when the user asks for them or their choice is genuinely unavailable (and the user still decides).
• Never fabricate model names — only recommend models listed in the CDN catalogs or returned by CLI.
• Never recommend a model outside the user's billing scope — Token Plan keys must only receive models from the Token Plan catalog linked above; PAYG keys may use the full catalog. Violating this causes hard failures for the user.
• Never invent or guess any price figure — use CLI / pricing.md / official pricing page only. Fabricating a price is a critical failure.
• Never silently fall back to snapshots when CLI errors out — apply error-handling.md recovery actions first.
• Never assume free quota is available — quotas may have been consumed, expired, or removed. Always present the paid unit price first.
• Never output API keys in plaintext — see Security section.
• Never confuse CLI session with API key — CLI auth uses browser device-flow login; never offer $DASHSCOPE_API_KEY or $QWEN_API_KEY as a fix for CLI Not authenticated / AUTH_REQUIRED errors.
• Never proactively fetch arbitrary URLs or trigger web searches. Fetch the documented CDN model catalogs when model data is needed; access other online sources only when CLI + CDN catalogs cannot answer AND the user confirms.
• Never construct usage/billing/console URLs — only use the exact links listed in this skill or its references. If a URL is not listed, do not invent one.
• Always include the cost disclaimer for any cost-related answer (see pricing-disclaimer.md).

References