Skip to trial filters
~/agentstafferAgentStaffer

Find, trial, compare, and govern AI agents before hiring

Trials Trial safely Next best action: Run a sandbox trial
Trials

Trial work queue

A compact sandbox trial queue for choosing a safe first task, checking fit and trust, and opening the full review package before any local run. Use this before hiring community-sourced agents into production workflows.

14available profiles
Filter optionsRole, protocol, trust, setup, and sorting stay available here.
First queue itemsglangReview trust and setup, then run a small code review trial.Review preflight
Setup coverage12 guided / 0 manualAcross 12 candidates on this page. A setup path is not safety approval; trust review and a sandbox trial are still required.Setup guide
Recent evidence0 runsNew sandbox evidence appears in the run ledger.Evidence queue
#1 trial priority · mcpservers / Coding & Development
Fit
61 / 100
Trust
Needs review
Setup
MCP ready
Before run
1 check
Top risk

No community evaluationAgentStaffer users have not submitted a review yet.

Full readiness snapshotCandidate summary and risk context stay available here.

Fast structured generation and serving for LLMs with RadixAttention prefix caching. Use for JSON/regex outputs, constrained decoding, agentic workflows with…

Top risk

No community evaluation: AgentStaffer users have not submitted a review yet.

Trial task Run a small code review trial
Input
Give the agent one low-risk issue or one small pull request and ask for defects, tests, and a patch plan.
Expected output
A concise review with reproducible findings, changed-file scope, and test suggestions.
Permissions & setup
  • Sandbox workspace: Run the first trial in a throwaway directory or isolated environment.
  • MCP-capable client: Use Claude Desktop, Cursor, VS Code, or another client that can load MCP server configs.
Promotion gates
  • Clear Install path reviewed
  • Hold Trust evidence reviewed
Guardrails
  • Install asks for production credentials or broad write access.
  • Source or package cannot be verified from profile evidence.

Create evidence with POST /api/actions/run-trial. Keep the card short; open the full checklist in the trial JSON.

Review & run trial Open profile
Review links
Profile JSON Trial JSON Setup JSON Trust JSON
#2 trial priority · playbooks / Coding & Development
Fit
61 / 100
Trust
Needs review
Setup
MCP ready
Before run
1 check
Top risk

No community evaluationAgentStaffer users have not submitted a review yet.

Full readiness snapshotCandidate summary and risk context stay available here.

This skill speeds up structured generation and agent workflows with prefix caching for JSON outputs and tool calls, boosting latency and throughput.

Top risk

No community evaluation: AgentStaffer users have not submitted a review yet.

Trial task Run a small code review trial
Input
Give the agent one low-risk issue or one small pull request and ask for defects, tests, and a patch plan.
Expected output
A concise review with reproducible findings, changed-file scope, and test suggestions.
Permissions & setup
  • Sandbox workspace: Run the first trial in a throwaway directory or isolated environment.
  • MCP-capable client: Use Claude Desktop, Cursor, VS Code, or another client that can load MCP server configs.
Promotion gates
  • Clear Install path reviewed
  • Hold Trust evidence reviewed
Guardrails
  • Install asks for production credentials or broad write access.
  • Source or package cannot be verified from profile evidence.

Create evidence with POST /api/actions/run-trial. Keep the card short; open the full checklist in the trial JSON.

Review & run trial Open profile
Review links
Profile JSON Trial JSON Setup JSON Trust JSON
#3 trial priority · playbooks / Workflow Automation
Fit
61 / 100
Trust
Needs review
Setup
MCP ready
Before run
1 check
Top risk

No community evaluationAgentStaffer users have not submitted a review yet.

Full readiness snapshotCandidate summary and risk context stay available here.

This skill helps you accelerate RL-based LLM post-training with slime's Megatron-LM and SGLang for scalable data generation and rollout.

Top risk

No community evaluation: AgentStaffer users have not submitted a review yet.

Trial task Run a dry-run workflow trial
Input
Connect the workflow to mock inputs and ask it to produce a trace without touching production systems.
Expected output
A step-by-step run log with inputs, outputs, failures, and rollback notes.
Permissions & setup
  • Sandbox workspace: Run the first trial in a throwaway directory or isolated environment.
  • MCP-capable client: Use Claude Desktop, Cursor, VS Code, or another client that can load MCP server configs.
Promotion gates
  • Clear Install path reviewed
  • Hold Trust evidence reviewed
Guardrails
  • Install asks for production credentials or broad write access.
  • Source or package cannot be verified from profile evidence.

Create evidence with POST /api/actions/run-trial. Keep the card short; open the full checklist in the trial JSON.

Review & run trial Open profile
Review links
Profile JSON Trial JSON Setup JSON Trust JSON
#4 trial priority · playbooks / Coding & Development
Fit
61 / 100
Trust
Needs review
Setup
MCP ready
Before run
1 check
Top risk

No community evaluationAgentStaffer users have not submitted a review yet.

Full readiness snapshotCandidate summary and risk context stay available here.

This skill helps you deploy and optimize high-throughput LLM inference with OpenAI-compatible APIs, reducing latency and enabling scalable multimodal models.

Top risk

No community evaluation: AgentStaffer users have not submitted a review yet.

Trial task Run a small code review trial
Input
Give the agent one low-risk issue or one small pull request and ask for defects, tests, and a patch plan.
Expected output
A concise review with reproducible findings, changed-file scope, and test suggestions.
Permissions & setup
  • Sandbox workspace: Run the first trial in a throwaway directory or isolated environment.
  • MCP-capable client: Use Claude Desktop, Cursor, VS Code, or another client that can load MCP server configs.
Promotion gates
  • Clear Install path reviewed
  • Hold Trust evidence reviewed
Guardrails
  • Install asks for production credentials or broad write access.
  • Source or package cannot be verified from profile evidence.

Create evidence with POST /api/actions/run-trial. Keep the card short; open the full checklist in the trial JSON.

Review & run trial Open profile
Review links
Profile JSON Trial JSON Setup JSON Trust JSON
#5 trial priority · mcpservers / Coding & Development
Fit
61 / 100
Trust
Needs review
Setup
MCP ready
Before run
1 check
Top risk

No community evaluationAgentStaffer users have not submitted a review yet.

Full readiness snapshotCandidate summary and risk context stay available here.

Provides guidance for LLM post-training with RL using slime, a Megatron+SGLang framework. Use when training GLM models, implementing custom data generation…

Top risk

No community evaluation: AgentStaffer users have not submitted a review yet.

Trial task Run a small code review trial
Input
Give the agent one low-risk issue or one small pull request and ask for defects, tests, and a patch plan.
Expected output
A concise review with reproducible findings, changed-file scope, and test suggestions.
Permissions & setup
  • Sandbox workspace: Run the first trial in a throwaway directory or isolated environment.
  • MCP-capable client: Use Claude Desktop, Cursor, VS Code, or another client that can load MCP server configs.
Promotion gates
  • Clear Install path reviewed
  • Hold Trust evidence reviewed
Guardrails
  • Install asks for production credentials or broad write access.
  • Source or package cannot be verified from profile evidence.

Create evidence with POST /api/actions/run-trial. Keep the card short; open the full checklist in the trial JSON.

Review & run trial Open profile
Review links
Profile JSON Trial JSON Setup JSON Trust JSON
#6 trial priority · claudemarketplaces / Coding & Development
Fit
44 / 100
Trust
Needs review
Setup
Source guide
Before run
1 check
Top risk

No community evaluationAgentStaffer users have not submitted a review yet.

Full readiness snapshotCandidate summary and risk context stay available here.

If you're building agents or chat systems with LLMs, this wrapper gives you structured outputs and automatic prefix caching through SGLang's RadixAttention. The real win is when you have repeated prefixes like system prompts or tool definitions. It'll cache them automatically and reportedly decode JSON about 3× faster than standard approaches. It's overkill...

Top risk

No community evaluation: AgentStaffer users have not submitted a review yet.

Trial task Run a small code review trial
Input
Give the agent one low-risk issue or one small pull request and ask for defects, tests, and a patch plan.
Expected output
A concise review with reproducible findings, changed-file scope, and test suggestions.
Permissions & setup
  • Sandbox workspace: Run the first trial in a throwaway directory or isolated environment.
  • License review: License metadata is missing from the current public profile.
Promotion gates
  • Clear Install path reviewed
  • Hold Trust evidence reviewed
Guardrails
  • Install asks for production credentials or broad write access.
  • Source or package cannot be verified from profile evidence.

Create evidence with POST /api/actions/run-trial. Keep the card short; open the full checklist in the trial JSON.

Review & run trial Open profile
Review links
Profile JSON Trial JSON Setup JSON Trust JSON
Show 6 more trial candidatesAdditional candidates stay folded for scanning; trials.json and card resource links keep the full review path.
#7 trial priority · mcpservers / Workflow Automation
Fit
60 / 100
Trust
Needs review
Setup
MCP ready
Before run
1 check
Top risk

No community evaluationAgentStaffer users have not submitted a review yet.

Full readiness snapshotCandidate summary and risk context stay available here.

Diagnose common DGX Station GB300 issues — CUDA crashes, wrong-GPU targeting, vLLM/SGLang container bugs, MIG state problems, NVLink/Fabric Manager errors,…

Top risk

No community evaluation: AgentStaffer users have not submitted a review yet.

Trial task Run a dry-run workflow trial
Input
Connect the workflow to mock inputs and ask it to produce a trace without touching production systems.
Expected output
A step-by-step run log with inputs, outputs, failures, and rollback notes.
Permissions & setup
  • Sandbox workspace: Run the first trial in a throwaway directory or isolated environment.
  • MCP-capable client: Use Claude Desktop, Cursor, VS Code, or another client that can load MCP server configs.
Promotion gates
  • Clear Install path reviewed
  • Hold Trust evidence reviewed
Guardrails
  • Install asks for production credentials or broad write access.
  • Source or package cannot be verified from profile evidence.

Create evidence with POST /api/actions/run-trial. Keep the card short; open the full checklist in the trial JSON.

Review & run trial Open profile
Review links
Profile JSON Trial JSON Setup JSON Trust JSON
#8 trial priority · claudemarketplaces / Coding & Development
Fit
42 / 100
Trust
Needs review
Setup
Source guide
Before run
1 check
Top risk

No community evaluationAgentStaffer users have not submitted a review yet.

Full readiness snapshotCandidate summary and risk context stay available here.

SGLang is a serving framework that's worth knowing about if you're building agents or need structured outputs. The killer feature is RadixAttention, which automatically caches shared prefixes across requests. In practice, this means if you're running agents with the same system prompt, you get 5× faster inference than vLLM because it only computes the new t...

Top risk

No community evaluation: AgentStaffer users have not submitted a review yet.

Trial task Run a small code review trial
Input
Give the agent one low-risk issue or one small pull request and ask for defects, tests, and a patch plan.
Expected output
A concise review with reproducible findings, changed-file scope, and test suggestions.
Permissions & setup
  • Sandbox workspace: Run the first trial in a throwaway directory or isolated environment.
  • License review: License metadata is missing from the current public profile.
Promotion gates
  • Clear Install path reviewed
  • Hold Trust evidence reviewed
Guardrails
  • Install asks for production credentials or broad write access.
  • Source or package cannot be verified from profile evidence.

Create evidence with POST /api/actions/run-trial. Keep the card short; open the full checklist in the trial JSON.

Review & run trial Open profile
Review links
Profile JSON Trial JSON Setup JSON Trust JSON
#9 trial priority · claudemarketplaces / Coding & Development
Fit
42 / 100
Trust
Needs review
Setup
Source guide
Before run
1 check
Top risk

No community evaluationAgentStaffer users have not submitted a review yet.

Full readiness snapshotCandidate summary and risk context stay available here.

This is THUDM's production RL framework that powers the GLM-4 series, built for teams who need Megatron-LM's native training with SGLang's fast rollout generation. If you're scaling post-training on GLM, Qwen3, or DeepSeek models and want tight control over the data buffer and generation workflow, this is your option. It has both sync and async modes, suppo...

Top risk

No community evaluation: AgentStaffer users have not submitted a review yet.

Trial task Run a small code review trial
Input
Give the agent one low-risk issue or one small pull request and ask for defects, tests, and a patch plan.
Expected output
A concise review with reproducible findings, changed-file scope, and test suggestions.
Permissions & setup
  • Sandbox workspace: Run the first trial in a throwaway directory or isolated environment.
  • License review: License metadata is missing from the current public profile.
Promotion gates
  • Clear Install path reviewed
  • Hold Trust evidence reviewed
Guardrails
  • Install asks for production credentials or broad write access.
  • Source or package cannot be verified from profile evidence.

Create evidence with POST /api/actions/run-trial. Keep the card short; open the full checklist in the trial JSON.

Review & run trial Open profile
Review links
Profile JSON Trial JSON Setup JSON Trust JSON
#10 trial priority · claudemarketplaces / Coding & Development
Fit
42 / 100
Trust
Needs review
Setup
Source guide
Before run
1 check
Top risk

No community evaluationAgentStaffer users have not submitted a review yet.

Full readiness snapshotCandidate summary and risk context stay available here.

This handles reinforcement learning training for LLMs using ByteDance's verl framework, which powers their Doubao models. It supports GRPO, PPO, and other RL algorithms with flexible backends that let you swap between FSDP, Megatron-LM, vLLM, and SGLang. The skill walks through math reasoning workflows, critic-based PPO setups, and large-scale training conf...

Top risk

No community evaluation: AgentStaffer users have not submitted a review yet.

Trial task Run a small code review trial
Input
Give the agent one low-risk issue or one small pull request and ask for defects, tests, and a patch plan.
Expected output
A concise review with reproducible findings, changed-file scope, and test suggestions.
Permissions & setup
  • Sandbox workspace: Run the first trial in a throwaway directory or isolated environment.
  • License review: License metadata is missing from the current public profile.
Promotion gates
  • Clear Install path reviewed
  • Hold Trust evidence reviewed
Guardrails
  • Install asks for production credentials or broad write access.
  • Source or package cannot be verified from profile evidence.

Create evidence with POST /api/actions/run-trial. Keep the card short; open the full checklist in the trial JSON.

Review & run trial Open profile
Review links
Profile JSON Trial JSON Setup JSON Trust JSON
#11 trial priority · claudemarketplaces / Coding & Development
Fit
34 / 100
Trust
Needs review
Setup
Source guide
Before run
1 check
Top risk

No community evaluationAgentStaffer users have not submitted a review yet.

Full readiness snapshotCandidate summary and risk context stay available here.

Production fork of slime for enterprise RL training on massive MoE models. The real draw here is unified FP8 training and inference with bit-wise alignment between SGLang rollouts and Megatron training, which prevents the routing inconsistencies that kill MoE stability. Supports INT4 quantization-aware training (fit 1TB models on single H200s) and speculati...

Top risk

No community evaluation: AgentStaffer users have not submitted a review yet.

Trial task Run a small code review trial
Input
Give the agent one low-risk issue or one small pull request and ask for defects, tests, and a patch plan.
Expected output
A concise review with reproducible findings, changed-file scope, and test suggestions.
Permissions & setup
  • Sandbox workspace: Run the first trial in a throwaway directory or isolated environment.
  • License review: License metadata is missing from the current public profile.
Promotion gates
  • Clear Install path reviewed
  • Hold Trust evidence reviewed
Guardrails
  • Install asks for production credentials or broad write access.
  • Source or package cannot be verified from profile evidence.

Create evidence with POST /api/actions/run-trial. Keep the card short; open the full checklist in the trial JSON.

Review & run trial Open profile
Review links
Profile JSON Trial JSON Setup JSON Trust JSON
#12 trial priority · smithery / General Agent
Fit
24 / 100
Trust
Needs review
Setup
Source guide
Before run
1 check
Top risk

No community evaluationAgentStaffer users have not submitted a review yet.

Full readiness snapshotCandidate summary and risk context stay available here.

Sglang is a Smithery skill listing by davila7.

Top risk

No community evaluation: AgentStaffer users have not submitted a review yet.

Trial task Run a sandbox capability trial
Input
Ask the agent to complete one narrow task that reflects your intended use, using public or mock data only.
Expected output
A useful result, clear assumptions, visible limitations, and enough evidence for a human evaluation.
Permissions & setup
  • Sandbox workspace: Run the first trial in a throwaway directory or isolated environment.
  • License review: License metadata is missing from the current public profile.
Promotion gates
  • Clear Install path reviewed
  • Hold Trust evidence reviewed
Guardrails
  • Install asks for production credentials or broad write access.
  • Source or package cannot be verified from profile evidence.

Create evidence with POST /api/actions/run-trial. Keep the card short; open the full checklist in the trial JSON.

Review & run trial Open profile
Review links
Profile JSON Trial JSON Setup JSON Trust JSON

Recent Trial Runs

RunResultScoreRolePromotionDetails
No public trial runs have been created yet. Pick a candidate above, open its profile, then use Run Trial or POST /api/actions/run-trial with the candidate listingId. New evidence appears here and in the trial guide.
Review support

Review links

Trial candidates and recent runs stay first; complete review records and finalist profile resources stay linked here.

Open Trial queue JSON
Supporting review linksLinks, review order, and page-specific notes remain available here.
Trial queue JSONEvidence requests JSONPromotion queue JSONGEO exposure JSONAgent visibility JSONResource catalog JSONSite context TXTSitemap XMLProfile package JSON
Trial review paths
Trial review path
  1. 01Site contextShared review rules
  2. 02Trial queueOpen the focused review record
  3. 03Profile pageOpen the candidate page
  4. 04Profile resourcesEvidence, setup, trust, trial, and history files

Use the links and review path below for trial queue context, evidence requests, and finalist profile resources before action.

Review route: OpenAI review tools can start with openai.json; otherwise use site context and agent inventory before opening the finalist's complete review materials before install, hire, or promotion.