Borrowing it
Nothing to install: this file belongs to tenstorrent/tt-studio. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/tenstorrent/tt-studio/main/.claude/skills/model-bringup/SKILL.mdgit clone --depth 1 https://github.com/tenstorrent/tt-studioWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/tenstorrent/tt-studio/model-bringup)<a href="https://agentmods.dev/skills/tenstorrent/tt-studio/model-bringup"><img src="https://agentmods.dev/badge/skills/tenstorrent/tt-studio/model-bringup/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/tenstorrent/tt-studio/model-bringup"><img src="https://agentmods.dev/badge/skills/tenstorrent/tt-studio/model-bringup.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00144 | $0.03236 |
| Opus 5 | $0.00072 | $0.01618 |
| Sonnet 5 | $0.00029 | $0.00647 |
| Haiku 4.5 | $0.00014 | $0.00324 |
Grade C, and why
model-bringup scanned grade C with 3 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Sends data to an external URLmediumData exfiltration
A POST to an outside endpoint may be telemetry or may be exfiltration; either way the mod talks to somewhere, and you should know where.
curl -X POST :8000/docker/deploy/ -d '{"model_id": "<id>", "weights_id": ""}' Asks for rootmediumPrivilege escalation
A mod that escalates privileges can change anything on the machine, not only the project.
| Token updated via Settings UI but container still gets the old one | **tt-studio** (known bug #1235) | `user_config.env` in the persistent volume is root-owned; the host inference-api can't read it and silently falls b Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -s :8000/docker/get_containers/ | grep <model-name> How it starts
The opening of the file, as written. The whole thing — 219 lines — stays where its author put it; the contents beside it link to each section on GitHub.
TT-Studio Model Bring-up
Battle-tested playbook for taking a model from "listed in tt-inference-server" to "generating output through the TT-Studio UI", including the cross-repo triage when it doesn't work. Motif-Image-6B-Preview on p300x2 (tt-studio#1234, tt-metal#53197, tt-inference-server#4955/#4956) and mochi-1-preview are the reference runs of this playbook.
Ground rules
- Work in a git worktree off
origin/dev(git worktree add .claude/worktrees/<name> -b <username>/<feature> origin/dev). Never edit the main checkout's branch — it may be another session's live work. - All git/PR mechanics follow the feature-branch-pr skill: branch
<username>/<feature>off dev, minimal in-scope diff, human commit messages, PR targetsdev, no AI attribution anywhere (no Co-Authored-By trailers, no mention of AI tools in commits, PR text, or issues). - Ask the user before stopping someone's running deployment or changing shared
state (tokens,
.env, board resets).
1. Recon before touching anything
- Upstream truth: read the model_support doc
(
tt-inference-server/docs/model_support/<type>/<Model>_<device>.md) — HF repo, docker image + tag, impl, inference engine, tt-metal commit. Treat its "Complete 🟢" status as a claim to verify, not a fact: the Motif doc's own pinned image could not run the model at all. - Artifact spec (the deploy-time source of truth for prod models):
.artifacts/tt-inference-server/workflows/model_specs/<env>/*.yaml. If the model + device entry is already in the pinned artifact version, no artifact bump and norequires_dev_catalog/inference_artifact_refare needed. Rootmodel_spec.json/release_model_spec.jsonare generated exports — never edit them;.artifacts/edits are ephemeral anyway. - tt-studio catalog: check
app/backend/shared_config/models_from_inference_server.jsonfor the model and for a same-type sibling to copy field-for-field (e.g. FLUX.1-dev for an image model). Field mapping rules live inapp/backend/shared_config/sync_models_from_inference_server.py(map_model_type,map_service_route,DEVICE_TYPE_TO_CONFIG). - Hardware reality:
tt-smi -ls,GET :8000/docker/chip-status/. Know the board semantics: a P300x2 "board" is ALL 4 chips (2 p300c cards); one multi-chip deployment occupies the whole board, and another multi-chip model cannot deploy until it's stopped. - What's running:
docker ps,GET :8000/docker/deployments/. If chips are occupied, ask the user before stopping anything, and record the occupant'smodel_idso it can be redeployed.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 219 lines · 144 tokens per session scan C 39f731fffe24
model-bringup is a skill published in the GitHub repository tenstorrent/tt-studio (50 stars, last pushed today), licensed Apache-2.0. It adds 144 tokens to every session and 3,236 once invoked, about $0.0007 per session on Opus 5. A static security scan graded it C with 3 findings (sends data to an external url, asks for root, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
langchain
Framework for building LLM-powered applications with agents, chains, and RAG. Supports multiple providers (OpenAI, Anthropic, Google), 500+ integrations, ReAct agents, tool calling, memory management, and vector store retrieval. Use for building chatbots, question-answering systems, autonomous agents, or RAG…
tamagui
Universal React UI framework for web and native. Use when building cross-platform apps with Tamagui, creating styled components with styled(), configuring design tokens/themes, using Tamagui UI components, or working with animations. Triggers: "tamagui", "styled()", "$token", "XStack/YStack", "useTheme", "@tamagui/"…
thinking-orbs
Add accessible animated AI loading and agent-status indicators with the React thinking-orbs library. Use when a chat, copilot, voice, search, generation, or tool-running interface needs a semantic working, searching, solving, listening, composing, or shaping state; when replacing a generic spinner with an AI activity…
git-safety
Git safety rules. INVOKE WHEN: git push, force push, git reset, git clean, destructive git, push force, reset hard. NEVER force push or do destructive git operations.
release-safety
Release safety rules. INVOKE WHEN: yarn release, npm publish, release canary, release packages, publishing, skip checks, skip tests. NEVER skip checks or tests without explicit permission.
assistant-ui
Guide for assistant-ui library - AI chat UI components. Use when asking about architecture, debugging, or understanding the codebase.