Run Command Builder
Configure and update commands shown in a VS Code Agent Session's Run button.
The skill requires reading the existing tasks.json, merging rather than overwriting unrelated tasks, and disclosing worktree-triggered execution, providing basic scoping. However, inferred shell commands need no user confirmation, and the skill does not disclose command data flow, privilege or sensitive-data risks, or provide rollback and verification steps, so points are deducted.
The decision flow and tasks.json format are clear, and package.json exposes recognizable commands. However, there is no handling or diagnostic guidance for invalid JSON, missing directories, command failures, platform differences, or merge conflicts. This is a static review without skill-specific key-path tests, so the score is capped below 10.
The audience, trigger scenario, input file, and output location are fairly clear, and several project stacks are covered. Boundaries for non-standard or multi-command projects, platform variation, and cases where automatic installation is inappropriate are under-specified, and Chinese-language interaction is not addressed.
The name, description, sectioning, examples, and core behavior are readable. The README, MIT license, version, and Microsoft attribution provide repository-level governance context, but the skill itself lacks versioning, changelog, maintenance ownership, installation notes, known limitations, and troubleshooting guidance, so points are deducted.
For simple projects, it can produce directly usable tasks.json entries for the Run button and explain how to trigger them. Inference may still produce incomplete or unsuitable commands, automatic installation can have side effects, and there is no evidence validating generated results or representative outputs; the static calibration therefore limits this to 7 or less.
The skill provides auditable schema, decision logic, and examples, while package.json and general CI files offer contextual evidence. There are no tests, fixed input/output fixtures, or third-party execution records for this skill, so the conclusion relies mainly on static documentation and is capped below 5.
- Automatically inferred shell commands, especially worktreeCreated installation commands, may perform network access, scripts, and side effects without confirmation; add confirmation, command preview, privilege and sensitive-data warnings, and recovery guidance.
- Specify handling for platform-specific commands, multiple candidates, invalid or missing tasks.json files, and merge conflicts, and validate the generated JSON.
- Repository-level CI and licensing do not substitute for skill-specific tests, version history, or maintenance documentation.
What it does & when to use it
This skill helps users configure run commands for the current Agent Session workspace. It represents those commands as tasks in .vscode/tasks.json so they appear in the session title bar's Run button. It checks for existing inAgents tasks before deciding whether the request is a modification or a new configuration. It is useful for projects that need consistent setup, development, or other workspace entry points.
Reads .vscode/tasks.json and checks for tasks with inAgents: true; if such tasks exist, asks whether the user wants to add, remove, or update a command; otherwise examines files such as package.json, Makefile, pyproject.toml, Cargo.toml, go.mod, and .nvmrc to infer setup and development commands; merges the resulting tasks into .vscode/tasks.json at the workspace root and can configure setup tasks to run on worktreeCreated.
- A project maintainer wants dependency installation to run automatically when a new worktree is created.
- A developer wants a development server command available from the VS Code Agent Session Run button.
- A team member needs to update one existing run command without replacing unrelated tasks.
- A developer working with standard project files wants clear setup and development commands inferred from the workspace.
Pros & cons
- Detects existing inAgents tasks and treats them as modification requests.
- Supports automatic setup tasks through worktreeCreated.
- Merges tasks.json while preserving unrelated tasks.
- Can infer commands from several common project configuration files.
- The source does not provide a standalone installation procedure or command.
- It asks the user to decide when the project structure or command choice is ambiguous.
- The source shows command generation and file writing, but does not establish that generated commands are executed or tested by the skill.
- The Run button and Agent Session behavior depend on the corresponding VS Code workspace environment.
How to install
The repository bundles 61 skills, and this skill is located at src/vs/sessions/skills/generate-run-commands/SKILL.md. The supplied README does not document a standalone installation command; the repository is licensed under MIT.
How to use
In a project-backed Agent Session workspace, ask for a request such as “Set up run commands for this project” or “Update the development command in the Run button.” The skill first checks .vscode/tasks.json; after configuration, the generated tasks can be triggered from the Run button in the session title bar.