Developer Productivity

Builds a throwaway prototype to answer one design question: a clickable HTML model for logic, or several UI variations to compare.

Create, edit and validate draw.io diagram files with correct mxGraph XML: flowcharts, architecture, sequence, ER and UML class diagrams, with templates and helper scripts.

Debug a failing GitHub Actions check on a pull request: find the failing checks, pull the logs, summarize the cause, draft a fix plan, and change code only after you approve.

Review a codebase like a security researcher: trace user input to dangerous sinks, find injection, access-control and secrets problems, and get patches to review, never applied automatically.

Points the agent at the agent-browser CLI to automate a real Chrome: navigate, fill forms, click, screenshot, extract data and test web apps.

Reads a code project and writes a step-by-step onboarding doc set (architecture, design rationale, code walkthrough, build, debugging) into docs/.

Evidence before claims: the model may not say work is done, fixed or passing until it has run the verification command in the same turn and read the output.

Makes your agent find and prove the cause of a bug before it changes any code.

Measure first, then fix: Core Web Vitals targets, a measure-identify-fix-verify loop, diagnosis for slow LCP, INP and CLS, backend N+1 and pagination patterns, and a list of anti-patterns, with a bundled checklist.

Hand a finished piece of work to a fresh reviewer: a code-reviewer prompt with the plan and the exact git range, graded findings, and rules for acting on them.

Turn a spec into a step-by-step implementation plan another engineer can follow: exact files, interfaces, test-first steps with expected output and a self-review.

Start a relentless interview that sharpens a plan or design, one question at a time.

Interview you about a plan or design and write down the decisions as ADRs and a glossary along the way.

Scan a codebase for refactors that turn shallow modules into deep ones, present them as a visual HTML report with before and after diagrams, then walk through whichever one you pick in a questioning loop.

Official Redis guidance on picking the right data structure (String, Hash, List, Set, Sorted Set, JSON, Stream, Vector Set) and naming keys consistently.

Helps the agent discover and install agent skills when you ask "is there a skill for X?".

Stop the agent from cutting corners: no placeholder code, no "rest follows the same pattern", no skipped sections. It counts the deliverables, writes every one in full and pauses cleanly at a token limit.

Move issues and external PRs through a small state machine of triage roles: categorise, verify, grill if needed, and write agent-ready briefs.

Google's workflows for making BigQuery cheaper and faster: SQL rewrites, partitioning and clustering, storage billing, editions and acceleration structures.

Security-first coding practice: threat model first, a three-tier Always / Ask first / Never boundary, and concrete controls for input, auth, sessions, headers, SSRF, secrets, dependencies, privacy and LLM features.

Review the changes since a commit, branch or tag along two separate axes: Standards (does it follow the repo's documented conventions and a baseline of common code smells) and Spec (does it do what the issue or spec asked), run in parallel and reported side by side.

Give an AI coding agent the right information at the right time: a rules file, the relevant spec and source, trimmed error output and a managed conversation, with packing strategies and rules for handling confusion.

Google's guidance for adding Sign In With Google to a web app: backend-mediated architectures first, ID token verification, CSP/COOP headers and sign-out.

Official Redis guidance for hardening a production instance: authentication and TLS, least-privilege ACL users, network exposure and dangerous commands.