Dev & Coding

A disciplined Git operator for atomic commits, safe rebases, and evidence-backed history investigation.

A rule-driven Git agent that splits work into atomic commits, performs safe rebases/squashes, and digs through history.

Automate web navigation, form filling, screenshots and scraping through small scripts that share persistent browser page state.

A Git expert skill that splits changes into atomic commits, drives rebase/squash safely, and digs through history with blame, bisect and pickaxe search.

A Scrum toolkit for writing user stories, acceptance criteria, splitting epics, planning sprint capacity, and prioritizing the backlog.

Analyzes every open GitHub issue and PR in parallel and writes permalink-backed triage reports without ever mutating the repository.

Inspects a GitHub PR, GitLab MR, or Perforce changelist for review comments, failing checks and description gaps, then helps fix them.

x

Keeps watching a pull request — CI, comments, and unresolved review threads — until it's genuinely ready to merge.

Runs a long-horizon experiment loop that improves a code/model/prompt artifact against an evaluator, using a persistent hypothesis tree and a held-out test merge gate to avoid overfitting.

Unpacks an installed Electron app's app.asar, restoring original sources from .js.map files or Prettier-formatting minified code.

Drafts a technical design document section by section, with an explicit approval gate after each section, based on a brief from the brainstorm step.

Scans a repository and builds an interactive knowledge graph of its architecture, components, and relationships.

Guides Claude to write correct Astropy code for units, coordinate transforms, FITS I/O, tables, time scales, WCS, and cosmology.

Cross-references git diffs or PRs against your project's knowledge graph to report changed components, blast radius, and risk.

Explains a specific file, function, or module in depth by combining a pre-built knowledge graph with the actual source code.

Turns Claude into an explore-first planning consultant that grounds in your codebase, asks only the decisions you must own, and — only after your explicit okay — writes one decision-complete work plan.

A post-implementation review gate that spawns five parallel sub-agents — goal verification, hands-on QA, code quality, security, and context mining — and passes only if all five pass.

A four-language engineering constitution that forces reference reading before coding, then enforces strict types, TDD, and file-size discipline.

Scans a repository and auto-generates a hierarchical set of AGENTS.md files at the root and in complexity-scored subdirectories.