Browse Skills

Add temporary unconditional JSONL logging to observe runtime behavior, reproduce the issue, and pinpoint root cause from log evidence before fixing.

Capture a screenshot and element snapshot of the current browser page when an error appears during e2e testing. Use when an agent-browser snapshot shows a failure like 'Something went wrong', auth error, blank page, or unexpected redirect.

Diagnoses root causes, implements minimal fixes with red→green regression tests, and validates the solution. Use for fixing bugs, test failures, or unexpected behavior.

Stress-tests a plan, spec, or design by interviewing the author one question at a time until every open decision is resolved, cross-checking answers against the codebase and surfacing contradictions.

Generates a self-contained HTML review page that explains a PR's changes, context, and verification evidence. Use when reviewers need a clear, standalone summary of a complex pull request.

Splits a large or messy PR into smaller, reviewable stacked or parallel PRs without losing any work. Extracts changes non-interactively, verifies each PR independently, and tracks drift from review feedback.

Performs read-only root-cause analysis on bugs, test failures, or unexpected behavior, producing a structured brief with the root cause, files to change, approach, and risks. Use before proposing any fix.

A guide for writing idiomatic Rust code, covering ownership patterns, error handling, performance optimization, testing, and documentation. Use when writing new Rust code or reviewing existing code for best practices adherence.

Rust coding standards for Carbon's workspace: tokio discipline for C++ FFI, ownership patterns, error conventions, and performance profiling based on measured results. Use when writing or reviewing Rust code in this repo.

Critically review your own branch diff against main before opening or merging a PR. Produces must-fix items, risks, suggested improvements, and a docs-freshness assessment. Supports an optional strict 'thermo-nuclear' mode for deep maintainability and abstraction audit.

Performs a quick end-to-end smoke test of the Carbon ERP dev server by logging in and verifying each core module renders without errors. Use after deploying changes or when asked to smoke test the app.

Designs features collaboratively by researching competitors, identifying design decisions, and resolving open questions with stakeholders before writing a complete specification document. Ideal for multi-file or cross-module changes.