Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Provider Management Skill · aiskillstore | DecimalAI

aiskillstore/provider-management

SkillScore100Maintained

Skill for managing model provider priorities with authentication (OAuth/Subscription/API), usage limits, and automatic fallback across all major AI providers

agent-toolingby aiskillstore

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

AI Engineer pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/aiskillstore-provider-management/SKILL.md

decimalai skills pull aiskillstore-provider-management --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/aiskillstore-provider-management/SKILL.md https://app.decimal.ai/s/aiskillstore-provider-management/SKILL.md

Raw file: /s/aiskillstore-provider-management/SKILL.md ↗ · machine-readable summary at /s/aiskillstore-provider-management.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 21d ago

Yes — +77% lift+14% tokens0% turns100% of whole cases pass with skill (22/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-01	✗→✓	▲ Improved	12%	0%
case-02	✗→✓	▲ Improved	-20%	0%
case-03	✗→✓	▲ Improved	-27%	0%
case-04	✗→✓	▲ Improved	36%	0%
case-05	✗→✓	▲ Improved	42%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/10/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

aiskillstore / marketplace

skills/attendantlion8/provider-management/SKILL.md

View on GitHub ↗

Imported 8/10/2026Version v1

Skill contents

SKILL.md

Provider Management Skill

Comprehensive management of AI models across all major providers with multi-auth support and automatic fallback.

Overview

Supports ALL major AI providers and models with three authentication methods:

| Priority | Auth Type | Description | |----------|-----------|-------------| | 100 | Subscription | Pro/Plus/Max subscriptions (highest priority) | | 50 | OAuth | OAuth 2.0 tokens (Antigravity-style) | | 10 | API | Direct API keys (lowest priority) |

Supported Models

Claude (Anthropic)

claude-4-opus, claude-sonnet-4, claude-3.5-sonnet, claude-3.5-haiku, claude-3-opus
Providers: anthropic-subscription, anthropic-oauth, anthropic-api, bedrock, vertex, openrouter

GPT (OpenAI)

gpt-4.1, gpt-4.1-mini, gpt-4o, gpt-4o-mini, o1, o3-mini
Providers: openai-subscription, openai-oauth, openai-api, azure, openrouter

Gemini (Google)

gemini-2.5-pro, gemini-2.5-flash, gemini-2.0-flash, gemini-1.5-pro
Providers: google-subscription, google-oauth, google-api, vertex-google, openrouter

Grok (xAI)

grok-3, grok-3-mini
Providers: xai-subscription, xai-api, openrouter

Other Models

Llama: llama-4-maverick, llama-3.3-70b (bedrock, together, groq, openrouter)
DeepSeek: deepseek-r1, deepseek-v3 (deepseek-api, together, openrouter)
Mistral: mistral-large, codestral (mistral-api, bedrock, azure, openrouter)
Cohere: command-r-plus (cohere-api, bedrock, openrouter)
Qwen: qwen-2.5-72b (together, openrouter)

Provider-Specific Model IDs

| Canonical | anthropic-api | bedrock | vertex | openrouter | |-----------|---------------|---------|--------|------------| | claude-sonnet-4 | claude-sonnet-4-20250514 | anthropic.claude-sonnet-4-20250514-v1:0 | claude-sonnet-4@20250514 | anthropic/claude-sonnet-4 | | gpt-4o | gpt-4o-2024-11-20 | - | - | openai/gpt-4o | | gemini-2.5-pro | - | - | gemini-2.5-pro-preview-05-06 | google/gemini-2.5-pro-preview |

Available Commands

`/provider-auth`

Configure authentication for providers.

bash
# View all auth status
/provider-auth status

# Set API key
/provider-auth setup anthropic-api --key sk-ant-...

# Start OAuth flow (Antigravity-style)
/provider-auth oauth google-oauth

# Set subscription token
/provider-auth setup anthropic-subscription --key <session-token>

`/provider-models`

List models and provider mappings.

bash
# List all models
/provider-models list

# Filter by family
/provider-models list --filter claude

# Get model info with all provider IDs
/provider-models info claude-sonnet-4

# Filter by capability
/provider-models capability reasoning

`/provider-priority`

Manage provider order.

bash
/provider-priority
/provider-priority set anthropic-subscription,anthropic-api,bedrock
/provider-priority move bedrock 1

`/provider-limits`

Configure usage limits.

bash
/provider-limits
/provider-limits anthropic-api --daily 1M --monthly 10M

`/provider-status`

View usage dashboard.

bash
/provider-status
/provider-status bedrock
/provider-status --reset

`/provider-switch`

Manual control.

bash
/provider-switch bedrock
/provider-switch --auto on

Configuration Files

~/.opencode/provider-fallback/
├── config.json   # Priority, limits, settings
├── usage.json    # Usage tracking
├── auth.json     # Stored credentials (0600 permissions)
└── tokens.json   # OAuth tokens

Authentication Priority Flow

1. Request comes in for model (e.g., claude-sonnet-4)
2. Get all configured providers for that model's vendor
3. Sort by auth priority: subscription (100) > oauth (50) > api (10)
4. Filter by usage capacity
5. Select best available provider
6. Get provider-specific model ID
7. Make request with appropriate credentials

OAuth Setup (Antigravity-style)

For Google, Anthropic, or OpenAI OAuth:

bash
# 1. Start OAuth flow
/provider-auth oauth google-oauth

# 2. Enter client ID and secret when prompted
# 3. Browser opens for authorization
# 4. Callback server receives token
# 5. Tokens stored securely

OAuth tokens auto-refresh when within 60 seconds of expiry.

Best Practices

Prioritize Subscriptions

Configure subscription auth first (highest priority, often includes extra features)
Add OAuth as secondary (good for personal accounts)
Use API keys as fallback

Multi-Vendor Setup

bash
# Configure multiple vendors
/provider-auth setup anthropic-subscription --key <token>
/provider-auth setup openai-api --key sk-...
/provider-auth setup google-api --key AIza...

# System auto-selects best for each model family

For Reliability

Configure 2-3 auth methods per vendor
Set conservative limits (80% of actual)
Keep auto-switch enabled
OAuth tokens refresh automatically

Troubleshooting

OAuth Token Expired

Tokens refresh automatically. If manual refresh needed:

bash
/provider-auth refresh google-oauth

Subscription Not Detected

Ensure session token is valid and not expired:

bash
/provider-auth setup anthropic-subscription --key <new-token>

Wrong Auth Method Used

Check configured providers and priorities:

bash
/provider-auth status

The system always prefers: subscription > oauth > api

Illustrative Want to see what aiskillstore/provider-management would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

9/2/2026, 16:26 UTC · 22 cases

Lift

+77%

100% vs 23%

Agent turns

1.0 vs 1.0

Tokens

+14%

2,135.55 vs 1,870.09

fail → pass · 17pass → pass · 5

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-01 I'm trying to set up my Anthropic credentials so I have a direct key ready if my subscription runs out. What commands should I run to register my API key `sk-ant-test12345` and then inspect all my configured auth methods?	fail→pass	171,142	12,280	-93%	1	1	0%	1,866	2,091	+12%	0	0	—
case-02 I need to map the standard Sonnet 4 model name to its provider-specific IDs across AWS Bedrock, Google Vertex, and OpenRouter. Could you show me the exact command to get detailed provider info for `claude-sonnet-4` and also how to search models by reasoning features?	fail→pass	24,287	86,025	+254%	1	1	0%	3,139	2,500	-20%	0	0	—
case-03 I want my system to prefer my Anthropic subscription first, then failover to direct API, and finally Bedrock. How do I reorder my provider execution hierarchy for these three and ensure automatic switching is turned on?	fail→pass	44,950	46,383	+3%	1	1	0%	3,243	2,369	-27%	0	0	—
case-04 I want to check my overall API usage dashboard across all providers, and then clear out my accumulated usage tracking stats. How do I inspect the general dashboard and reset my tracking counters?	fail→pass	31,306	28,199	-10%	1	1	0%	1,480	2,019	+36%	0	0	—
case-05 I want to set up Google login authentication using OAuth tokens instead of static keys for my AI queries. What command initiates the Google OAuth authorization flow?	fail→pass	11,978	8,169	-32%	1	1	0%	1,376	1,958	+42%	0	0	—
case-06 My Google OAuth token seems stuck and isn't auto-refreshing properly before expiration. Which command manually forces a credential token refresh for google-oauth?	fail→pass	29,277	10,452	-64%	1	1	0%	1,883	1,846	-2%	0	0	—
case-07 What is the specific AWS Bedrock model identifier corresponding to the canonical model `claude-sonnet-4`?	fail→pass	9,927	6,764	-32%	1	1	0%	958	1,841	+92%	0	0	—
case-08 When sending queries through Google Vertex AI for `claude-sonnet-4`, what exact model string must be sent to the endpoint?	fail→pass	13,444	13,409	-0%	1	1	0%	1,585	1,949	+23%	0	0	—
case-09 What versioned API model ID is used directly with Anthropic's direct API endpoints for `claude-sonnet-4`?	fail→pass	7,812	9,611	+23%	1	1	0%	1,207	1,978	+64%	0	0	—
case-10 If a user has configured subscription credentials, direct API keys, and OAuth tokens for the same model vendor, in what order does the system evaluate and select these authentication methods?	fail→pass	21,548	8,019	-63%	1	1	0%	2,374	2,168	-9%	0	0	—
case-11 Where are stored authentication credentials maintained on disk, and what file permission mode should be enforced on that file for security?	fail→pass	18,908	6,100	-68%	1	1	0%	2,456	2,030	-17%	0	0	—
case-12 I need to shift AWS Bedrock to become my top-priority provider without retyping the entire list of providers. What command moves Bedrock directly to position 1?	fail→pass	13,477	7,833	-42%	1	1	0%	1,431	1,909	+33%	0	0	—
case-13 I need to set usage caps for my Anthropic direct API key to prevent overspending. How do I configure a daily quota of 1M tokens and a monthly quota of 10M tokens?	fail→pass	17,878	9,924	-44%	1	1	0%	2,919	2,043	-30%	0	0	—
case-14 I want to temporarily bypass the priority router and force all subsequent AI calls to route strictly through AWS Bedrock. Which command switches active routing to Bedrock?	fail→pass	16,410	10,674	-35%	1	1	0%	1,622	2,014	+24%	0	0	—
case-15 I want to list only the available Claude model variants rather than scrolling through models from all vendors. What CLI option filters the list by the claude family?	fail→pass	11,019	7,846	-29%	1	1	0%	1,782	1,893	+6%	0	0	—
case-16 I have an active subscription and extracted my session token. What command attaches this session token to the `anthropic-subscription` provider?	fail→pass	9,607	5,742	-40%	1	1	0%	1,818	1,924	+6%	0	0	—
case-17 How close to expiration does an OAuth access token need to be before the system automatically triggers a refresh cycle?	pass→pass	9,984	2,203	-78%	1	1	0%	1,823	1,870	+3%	0	0	—
case-18 What model slug is passed when invoking `gpt-4o` through OpenRouter?	pass→pass	8,278	3,784	-54%	1	1	0%	541	1,797	+232%	0	0	—
case-19 In which directory path are system provider configuration files, usage tracking logs, and token stores saved?	fail→pass	11,053	13,031	+18%	1	1	0%	2,115	1,899	-10%	0	0	—
case-20 How do I instantiate `anthropic.Anthropic()` in Python and handle `APIConnectionError` with a retrying exponential backoff loop using the standard `tenacity` library?	pass→pass	12,168	13,957	+15%	1	1	0%	2,192	3,540	+61%	0	0	—
case-21 Can you provide an AWS IAM JSON policy document that grants `bedrock:InvokeModel` permissions specifically for the `anthropic.claude-3-5-sonnet-20241022-v2:0` ARN?	pass→pass	9,109	5,002	-45%	1	1	0%	1,494	2,506	+68%	0	0	—
case-22 How do I create a Google Cloud GCP service account using gcloud CLI and assign it the `roles/aiplatform.user` role for Vertex AI access?	pass→pass	9,278	22,053	+138%	1	1	0%	1,838	2,838	+54%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +77 percentage points is the difference between those two pass rates over the 22 comparable cases.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 22.7%
Pass rate with the skill: 100%
Cases attempted: 22
Cases counted toward lift: 22
Cases that regressed: 0
Skill version measured: v1
Run completed: 9/2/2026, 16:26 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-01

Prompt

I'm trying to set up my Anthropic credentials so I have a direct key ready if my subscription runs out. What commands should I run to register my API key `sk-ant-test12345` and then inspect all my configured auth methods?

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run100

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns+59% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 20 days ago

Other measured skills in the registry, with their headline benchmark lift.