JetBrains/thinkrail is a desktop and mobile client that embeds the pi coding agent inside an interface for editing code and managing development workspaces. Developers use it to work on Git repositories through separate worktrees, a Monaco editor, terminals, Git views, specifications, and concurrent agent sessions. The catalogue entries are skills and instructions for its agent-based development workflow.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add JetBrains/thinkrail --skill starting-a-new-projectgit clone --depth 1 https://github.com/JetBrains/thinkrailWrote 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/jetbrains/thinkrail/starting-a-new-project)<a href="https://agentmods.dev/skills/jetbrains/thinkrail/starting-a-new-project"><img src="https://agentmods.dev/badge/skills/jetbrains/thinkrail/starting-a-new-project/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/jetbrains/thinkrail/starting-a-new-project"><img src="https://agentmods.dev/badge/skills/jetbrains/thinkrail/starting-a-new-project.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Excessive Agency · line 40 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
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 | $0.00061 | $0.01226 |
| Opus 5 | $0.00030 | $0.00613 |
| Sonnet 5 | $0.00012 | $0.00245 |
| Haiku 4.5 | $0.00006 | $0.00123 |
Grade A, and why
starting-a-new-project scanned grade A with 0 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 10d 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 89 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Starting a new project
The workspace is empty: no code, no decisions. Turn the user's idea into one clear, buildable document —
goal-and-requirements.md — then hand off to brainstorming for the features that follow.
Hold the writing-specs bar. Read that concept skill before saving anything — it carries the short / honest / on-rails rules every section you save must meet.
Method
- Build on what's already said. Never re-ask what the request already told you.
- Infer, then confirm — propose a concrete draft and let the user correct it; a suggestion beats an
open question. Compose
ask_user_questionrounds per the asking-user-questions concept skill (read it before the first round — it carries the option, confirmation, and degradation norms). - MVP first. The right v1 is smaller than the user expects. Every v1 capability must justify itself.
- Save incrementally. Create the file as soon as the first section is settled, then add each confirmed section in template order. Don't batch; don't invent unconfirmed content.
- A skipped question is not a blocker — proceed on the current model and note real gaps inline.
Working model (infer from the request; never ask these directly)
audience: personal | public | both domain: what space this is in
tech: stack mentioned, or null scope: small | large
depth: light | standard | full creator_is_user: does the maker use it?
depth scales the document: light = a one-liner idea → a few lines; full = named competitors /
multiple user types → a full PRD. It can only grow during the conversation, never shrink.
Fast path — pre-filled brief
If the request already reads like a spec (several headings or a multi-section brief), parse it, treat those
sections as confirmed, save them immediately, and only pursue what's genuinely missing and required by
depth. Don't ask the user to confirm what they already wrote. The one always-offered extra is
alternatives research (below).
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.
- 10d ago First seen · 89 lines · 61 tokens per session scan A bb3a3f740f32
starting-a-new-project is a skill published in the GitHub repository JetBrains/thinkrail (427 stars, last pushed yesterday), licensed Apache-2.0. It adds 61 tokens to every session and 1,226 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
file-a-task
File work into nohuman (taskadd) and check on it (taskstatus) via the nohuman MCP bridge, instead of doing the work inline.
vigilante-create-issue
Help a human author write an implementation-ready GitHub issue that Vigilante can execute reliably.
vigilante-issue-implementation
Implement a GitHub issue end-to-end when Vigilante dispatches work for a watched repository. Use the provided worktree, respect repository instructions, comment on the issue as work progresses, and report failures back to GitHub.
vigilante-issue-implementation-on-bazel-monorepo
Implement a GitHub issue end-to-end when Vigilante dispatches work for a Bazel-based monorepo. Use the provided worktree, respect repository instructions, comment on the issue as work progresses, and report failures back to GitHub.
vigilante-issue-implementation-on-rush-monorepo
Implement a GitHub issue end-to-end when Vigilante dispatches work for a watched Rush monorepo. Use the provided worktree, respect repository instructions, comment on the issue as work progresses, and report failures back to GitHub.
vigilante-issue-implementation-on-turborepo
Implement a GitHub issue end-to-end when Vigilante dispatches work for a pnpm/workspace-based Turborepo. Use the provided worktree, respect repository instructions, comment on the issue as work progresses, and report failures back to GitHub.