TensorPool GPU Cloud
Use this skill for TensorPool cluster, job, and storage workflows. Treat provider availability, instance names, quotas, and prices as live data: inspect them at execution time and never rely on values copied into a skill.
OpenScience credential boundary
- A key saved in Customize > Compute > TensorPool is encrypted control-plane data. It is not exported into Bash, Task, notebooks, kernels, plugins, or MCP servers.
- Use
provider_compute with account, list_resources, resource_status, list_jobs, or job_status for live reads. OpenScience selects the exact reviewed tp argv, admits the key only to that isolated child, and marks last_used only after success.
- Test connection uses the same boundary for exactly
tp --no-input me.
- The saved key is admitted only when Test connection approves a root/admin-managed, non-writable
tp executable. A normal user-owned pip, pipx, or Homebrew install is credential-only and cannot be used by provider_compute; use an administrator-managed install or wait for a managed/native adapter.
- Never print, echo, persist, or ask the user to paste a saved key into chat.
- A generic agent shell can use
tp only when the user has independently authenticated that shell. The saved credential cannot create, submit, edit, attach, destroy, or otherwise mutate resources; do not ask the user to weaken this boundary.
Current CLI contract
For an independently authenticated user shell, install and inspect the CLI rather than guessing its version. This does not make a user-owned install eligible for the saved-key broker:
python -m pip install tensorpool
tp --help
tp --no-input me
The first-party CLI reads TENSORPOOL_KEY. The read-only identity check is tp --no-input me.
The broker maps its read operations to these current official commands:
tp cluster list
tp job list
tp storage list
storage list is not yet exposed by the broker. Run it only in an independently authenticated user shell.
Before using any other command, run its current --help. CLI flags can change independently of this bundled skill.
Operating policy
- Confirm the intended workload, region or placement constraints, GPU count, runtime estimate, storage needs, and maximum budget.
- Inspect current inventory and provider-displayed cost immediately before proposing a resource.
- Show one recommended configuration and its current quoted cost. Label estimates as estimates.
- Get explicit user approval before cluster creation, job submission, resizing, attaching billable storage, or any other mutation.
- After launch, record the provider resource identifier, selected type, creation time, and the exact command used.
- Monitor to a terminal state. Download or verify required outputs before cleanup.
- Get confirmation before deleting persistent storage. Stop or delete idle compute promptly after the requested work finishes.
Never infer that an empty list means a provider outage; report the exact CLI result. Never retry a billable mutation blindly after a timeout—first list resources and reconcile whether the first request succeeded.
SSH handoff
When a cluster exposes SSH, add it through Customize > Compute > SSH. Prefer a pinned host key, a literal private-key path with restrictive permissions, and a validated ProxyJump when required. OpenScience does not evaluate ProxyCommand or Match exec from user SSH config.
Sources of truth
- CLI and product documentation: https://docs.tensorpool.dev/
- Installed command contract:
tp --help and the relevant subcommand --help
- Current inventory and pricing: TensorPool's live dashboard or CLI output at the time of approval
1---2name: tensorpool-gpu-cloud3description: Safely inspect and operate TensorPool GPU clusters and jobs using the current tp CLI, with explicit approval before any billable or destructive action.4license: MIT5---67# TensorPool GPU Cloud89Use this skill for TensorPool cluster, job, and storage workflows. Treat provider availability, instance names, quotas, and prices as live data: inspect them at execution time and never rely on values copied into a skill.1011## OpenScience credential boundary1213- A key saved in **Customize > Compute > TensorPool** is encrypted control-plane data. It is not exported into Bash, Task, notebooks, kernels, plugins, or MCP servers.14- Use `provider_compute` with `account`, `list_resources`, `resource_status`, `list_jobs`, or `job_status` for live reads. OpenScience selects the exact reviewed `tp` argv, admits the key only to that isolated child, and marks `last_used` only after success.15- **Test connection** uses the same boundary for exactly `tp --no-input me`.16- The saved key is admitted only when **Test connection** approves a root/admin-managed, non-writable `tp` executable. A normal user-owned pip, pipx, or Homebrew install is credential-only and cannot be used by `provider_compute`; use an administrator-managed install or wait for a managed/native adapter.17- Never print, echo, persist, or ask the user to paste a saved key into chat.18- A generic agent shell can use `tp` only when the user has independently authenticated that shell. The saved credential cannot create, submit, edit, attach, destroy, or otherwise mutate resources; do not ask the user to weaken this boundary.1920## Current CLI contract2122For an independently authenticated user shell, install and inspect the CLI rather than guessing its version. This does not make a user-owned install eligible for the saved-key broker:2324```bash25python -m pip install tensorpool26tp --help27tp --no-input me28```2930The first-party CLI reads `TENSORPOOL_KEY`. The read-only identity check is `tp --no-input me`.3132The broker maps its read operations to these current official commands:3334```bash35tp cluster list36tp job list37tp storage list38```3940`storage list` is not yet exposed by the broker. Run it only in an independently authenticated user shell.4142Before using any other command, run its current `--help`. CLI flags can change independently of this bundled skill.4344## Operating policy45461. Confirm the intended workload, region or placement constraints, GPU count, runtime estimate, storage needs, and maximum budget.472. Inspect current inventory and provider-displayed cost immediately before proposing a resource.483. Show one recommended configuration and its current quoted cost. Label estimates as estimates.494. Get explicit user approval before cluster creation, job submission, resizing, attaching billable storage, or any other mutation.505. After launch, record the provider resource identifier, selected type, creation time, and the exact command used.516. Monitor to a terminal state. Download or verify required outputs before cleanup.527. Get confirmation before deleting persistent storage. Stop or delete idle compute promptly after the requested work finishes.5354Never infer that an empty list means a provider outage; report the exact CLI result. Never retry a billable mutation blindly after a timeout—first list resources and reconcile whether the first request succeeded.5556## SSH handoff5758When a cluster exposes SSH, add it through **Customize > Compute > SSH**. Prefer a pinned host key, a literal private-key path with restrictive permissions, and a validated ProxyJump when required. OpenScience does not evaluate `ProxyCommand` or `Match exec` from user SSH config.5960## Sources of truth6162- CLI and product documentation: <https://docs.tensorpool.dev/>63- Installed command contract: `tp --help` and the relevant subcommand `--help`64- Current inventory and pricing: TensorPool's live dashboard or CLI output at the time of approval
synthetic-sciences