Claude Skills by badmuriss

Run durable task graphs when independent work needs ownership, integration checks and recovery. Use for graph execution, not routine edits.

Inspect and operate local browser or desktop interfaces. Prefer Jev Ultrafast for supported browser-only goals when its required credentials are available; use Orca for desktop apps, unsupported browser interactions, screenshots, and accessibility-level control.

Verify rendered UI changes with reproducible screenshots and vision review on their supported platforms.

Adversarial interview about a plan until you reach shared understanding. use_when: \"grill me\", stress-test a plan or design. do_not_use_when: the plan must be checked against project docs, use grill-with-docs.

Grill a plan against the documented domain model, updating CONTEXT.md and ADRs inline. use_when: stress-test a plan against the project's own language and recorded decisions. do_not_use_when: there is no domain documentation, use grill-me.

Implement a requested change or approved plan, verify its outcomes, and report concrete blockers. Use graph execution only for independent work that needs durable coordination.

Extract content from documents when their format or layout prevents reliable direct reading. Verify reading order and tables before analysis.

Make a public repository README easier to understand and use. Apply when README structure or presentation is requested.

Inspect and remove requested metadata or invisible marks from user-owned files. Use when provenance cleanup is explicitly requested.

Research external questions and compare evidence with source attribution. Use for research or fact-checking, not facts already established in local code.

Audit an existing corpus of agent instructions for conflicts, duplication and unnecessary process. Use for a whole-corpus audit, not individual rule edits.

Collect live public web data through ScrapingDog when an API key is available. Use for search, scraping and dedicated public-data endpoints.

Plan a requested codebase change or architecture decision. Use when planning is requested; do not implement code.

Perform a strict, read-only maintainability review of a completed implementation diff, including complexity audits and test economy. Use after impl, before PR preparation, or when asked for a thermonuclear, harsh, deep code-quality, cyclomatic-complexity, concise-tests, redundant-tests, architecture, spaghetti, abstraction, or large-file review. Report evidence-backed findings; never edit code or turn subjective preferences into automatic blockers.

Review source comments and remove low-value comments that only narrate visible code while preserving rationale, constraints, safety context, public documentation, and tooling directives. Use when asked to trim code comments or remove dumb, redundant, obvious, excessive, or AI-generated comments. Do not run automatically after implementation.

Write clear technical documentation, commit messages and PR descriptions. Use unslop for standalone prose when it is installed.