dotfiles

Prepare a claim-centered adversarial review packet for independent review. Use when a consequential, disputed, high-uncertainty, or large-scope design, recommendation, plan, assumption, implementation conclusion, or PR conclusion must be tested against evidence, scope, falsification criteria, and residual risk; not for ordinary patch defect review or security scanning.

Run and interpret Rust test coverage with cargo llvm-cov through a bundled script. Use for crate or workspace coverage summaries, uncovered lines/functions, diff-focused coverage review, targeted Rust test-gap recommendations, or cargo llvm-cov troubleshooting.

Maintain this chezmoi repo's portable package manifest. Use when adding, updating, removing, or verifying standalone GitHub release CLI packages in home/.chezmoidata/portable-pkgs.yaml.

Edit the shared agent contract — the chezmoi-managed AGENTS.md / CLAUDE.md instruction files generated from home/.chezmoitemplates/agents/ partials. Use when adding, changing, or removing a durable behavior rule for the agents, or when asked to update AGENTS.md, CLAUDE.md, the shared partials, or a tool's contract residue.

Drive a live marimo notebook as a workspace: run Python in the same kernel the user does, inspect live notebook state, and commit durable notebook changes. Use when the user wants to start a marimo notebook or pair on an active marimo session.

Capture or refresh persisted project conventions. Use when the user wants to initialize AGENTS.md, project style, validation commands, or tool preferences so agents stop rediscovering rules each session.

Create a GitHub pull request for the current branch. Use when the user wants to open a PR, optionally specifying a target branch (defaults to 'main'). The PR title and body should describe the branch's net change relative to the target branch, not the branch's internal history.

Create or revise an agent skill. Use when the user wants to add a new skill, improve an existing skill, define trigger conditions, design SKILL.md structure, or decide whether a skill should include scripts, references, or examples.

External repository research workflow with a controlled checkout script. Use when local code, official docs, release notes, or web pages are insufficient and a task requires temporarily cloning or inspecting a third-party repository as read-only reference material.

Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.

Create a git commit when the user explicitly asks for one. Use when the user wants to commit changes, create a commit message, or group and stage related changes before committing.

Write or revise Markdown that is clear, compact, and markdownlint-friendly. Use when the user asks for README content, skill docs, notes, guides, or any Markdown file that should be well-structured and lint-clean.

Use when editing, formatting, testing, reviewing, or designing Rust code. Covers project-local policy precedence, Rust toolchain and formatter selection, MSRV-aware idioms, ownership-aware API design, correctness boundaries, unsafe code, lints, dependencies, and tests.

Work on Python projects with uv. Use for repositories with pyproject.toml, uv.lock, package directories, dependency groups, .venv environments, uv sync/run/add/remove/lock workflows, or project-level tests and validation.

Create, edit, run, format, and check standalone single-file Python scripts outside a Python project with uv. Use before creating or running PEP 723 or uv run --script scripts; skip third-party scripts that should be followed as-is.

Persist agent behavior preferences by updating durable prompt/config files before using long-term memory.