By mujtaba3B
Engineer Earnie, the engineering persona. Works through PR review feedback on his own PRs, reviews other authors' PRs, watches CodeRabbit and applies fixes, spikes the riskiest unknown before building, generates CodeRabbit config, and builds macOS Shortcuts. Skills: eng:pr-feedback, eng:review-pr, eng:pr-watcher, eng:spike, eng:coderabbit-config, eng:shortcut.
Generate or update a tailored .coderabbit.yaml for the current repository. Inspects the repo (languages, monorepo shape, generated/vendored dirs, CLAUDE.md/AGENTS.md conventions) and produces a config with sensible defaults plus repo-specific path_filters and path_instructions. Use when the user says "/eng:coderabbit-config", "coderabbit config", "coderabbit yaml", "set up coderabbit for this repo", "customize coderabbit", "generate .coderabbit.yaml", "tune coderabbit", or otherwise asks to configure CodeRabbit AI code review for a repo. Run from inside the target repo (cwd = repo root). Defaults to inferring from the repo so the user only answers 1-2 questions.
Use this skill whenever the user wants to work through review comments on a pull request they authored — addressing each comment, patching the code, and capturing a durable lesson so the same mistake is not made again. Trigger when the user says "address PR comments", "review comments", "CR comments", "my PR has comments", "CodeRabbit flagged", "what does this reviewer want", "respond to reviewer", "fix review feedback", "learn from PR feedback", or "/eng:pr-feedback". Use this even when the user only says "there are comments on my PR" — the skill covers both the triage and the fix-plus-learn loop.
Foreground watcher that pairs the main agent (dispatcher and fix-applier) with a passive polling subagent (sensor) to handle CodeRabbit feedback on a GitHub PR. The sensor blocks silently in one Agent call until CR posts a settled round of feedback, then returns a single JSON blob. The main agent classifies, applies fixes, runs tests, commits, pushes, and replies on the PR itself, then spawns the next sensor. Never merges; never resolves conversations; never pushes without passing tests; never touches files outside the PR's own diff; never parallelizes fixes. Use when asked to "watch the PR", "watch coderabbit", "pr watch", or invoked manually as `/eng:pr-watcher <PR_URL>` after /ship.
Review someone else's GitHub pull request (typically one opened by an autonomous agent like MuTwo/MuThree) and leave one well-structured review comment on it. Resolves the PR, locates the pr-review-toolkit review engine (offers to install it if missing), runs the diff through six toolkit review lenses (bugs/CLAUDE.md, silent failures, test gaps, type design, comment rot, simplification) plus a seventh design/blast-radius lens the main agent always applies, verifies the sharp findings against the actual repo rather than trusting the PR description, gives the user a quick chat summary, then always posts exactly ONE structured comment via gh pr comment (led by an @<author> mention so the PR's creator is notified) after a confirm gate on the wording. The run always ends in a posted, author-tagged comment regardless of verdict. Never merges, never pushes, never resolves conversations, never touches the assignee. Use when asked to "review this PR", "review the agent PR", "is this good to deploy?", "look at PR #N", "review MuThree's PR", or invoked as `/eng:review-pr` with a PR number or URL.
Create a macOS/iOS Shortcut programmatically from a spec, without the user hand-building it in Shortcuts.app. Authors the workflow as a plist, signs it into a .shortcut file, opens it for the one mandatory "Add Shortcut" click, then verifies it runs. Use this whenever you need to wire up a Shortcut for the user, especially a "Run Shell Script" bridge that lets a Shortcut, the menu bar, a hotkey, or Siri invoke a CLI or script. Trigger on "create a shortcut", "make a Shortcuts workflow", "add a Shortcut that runs my script", "bridge this CLI to Shortcuts", "wire X into Shortcuts.app", or any request to generate/build/install an Apple Shortcut from code.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Personal skills layered on top of gstack, discovered by Claude Code alongside gstack's own skills.
Skills are organized into persona plugins: each persona is a Claude Code plugin installed from this repo's own local marketplace, and its skills invoke namespaced as <persona>:<skill> (for example /pm:bug, /design:pencil-mockup).
git clone <this-repo> ~/dev/gstack-extensions
cd ~/dev/gstack-extensions
./bin/install
Then restart your Claude Code session. Skills become invokable as /pm:bug, /qa:browser, /eng:pr-feedback, /design:pencil-mockup, etc.
Four persona plugins (each a top-level dir with a .claude-plugin/plugin.json, a skills/ tree, and shared context):
pm/): product-manager persona. /pm:feature, /pm:bug, /pm:next-issue, /pm:first-principles. Turns ideas, bug reports, and "what next?" into well-structured GitHub issues, and reframes problems from first principles.qa/): manual-QA persona. /qa:browser, /qa:headless. Verifies one defined flow against the spec or mockup, in the browser (driving the gstack browse daemon, AI-comparing screenshots against Pencil mockups) or headless (capturing backend side effects).eng/): engineering persona. /eng:pr-feedback, /eng:review-pr, /eng:pr-watcher, /eng:spike, /eng:coderabbit-config, /eng:shortcut. Works PR feedback, reviews others' PRs, watches CodeRabbit, spikes risky unknowns, configures CodeRabbit, and builds macOS Shortcuts.design/): design persona. /design:pencil-mockup. The Pencil-native counterpart to gstack's HTML design skills: creates and updates .pen mockups on the canvas via the Pencil MCP.The persona name (Penny / Quincey / Earnie / Denise) lives in each plugin's description and README as a memory hook; you invoke by the short role prefix, not the name.
./bin/install registers this repo as a local Claude Code marketplace (the root .claude-plugin/marketplace.json lists every plugin) and installs each plugin from it. Claude Code namespaces a plugin's skills as <name>:<skill>, where <name> is the plugin's manifest name (not the directory name). Each skill lives in <plugin>/skills/<slug>/SKILL.md and resolves sibling files (shared/*.md, references/) relative to its plugin root.
Installing copies each plugin into ~/.claude/plugins/cache/gstack-extensions/<name>/<version>/, so the repo is not read live: re-run ./bin/install (or bin/gstack-extensions-upgrade) to refresh the cache from the working tree, then restart the session. Because the plugins live in the plugin cache rather than in ~/.claude/skills/, they never collide on the filesystem with gstack's own skills: gstack's loose /qa and this repo's /qa:browser coexist cleanly (the : is what separates them).
This repo lives outside ~/.claude/skills/gstack/, so gstack-upgrade never touches it.
Earlier versions symlinked each plugin dir into
~/.claude/skills/. That only ever loaded as<name>@skills-dirand refused to load when a name collided with an installed plugin, so the marketplace install above replaced it.bin/installstill sweeps any leftover symlinks from that layout.
Each skill checks on invocation whether this clone's main is behind origin/main (a TTL-gated git fetch, so it does not hammer the remote) and, if so, offers to upgrade. Accepting runs:
~/dev/gstack-extensions/bin/gstack-extensions-upgrade
which fast-forwards main and refreshes the installed plugins from the pulled source (uninstall+install, since claude plugin update no-ops while a plugin's version is unchanged). It refuses safely (and tells you why) if the clone is not on a clean main, so it never disrupts in-progress feature-branch work. Restart the session afterwards to load the refreshed skills. To upgrade by hand at any time:
cd ~/dev/gstack-extensions
git pull --ff-only # must be on a clean main
./bin/install # idempotent; refreshes the installed plugins
bin/gstack-extensions-update-check is the read-only check behind the prompt; it prints UPGRADE_AVAILABLE <n> <range> when behind and nothing otherwise.
./bin/uninstall
Uninstalls the four plugins and removes this repo's local marketplace (and sweeps any leftover symlinks from the old installer). Leaves gstack and any other marketplaces alone.
A new skill in an existing persona: create <plugin>/skills/<slug>/SKILL.md with valid frontmatter (name, description) and the standard "Update check (run first)" preamble (copy it from any existing skill). Re-run ./bin/install to refresh the plugin in the cache, then restart the session to register it. It invokes as /<plugin>:<slug>.
QA Quincey, the manual-QA persona. Verifies one defined flow against the spec or mockup, either in the browser (driving the gstack browse daemon and AI-comparing screenshots against Pencil mockups) or headless (capturing backend side effects). Skills: qa:browser, qa:headless.
PM Penny, the product-manager persona. Turns ideas, bug reports, and 'what should I work on next?' into well-structured GitHub issues, and reframes problems from first principles. Skills: pm:bug, pm:feature, pm:next-issue, pm:first-principles.
Designer Denise, the design persona. Creates and updates Pencil (.pen) mockups via the Pencil MCP, the Pencil-native counterpart to gstack's HTML-oriented design skills. Skills: design:pencil-mockup.
npx claudepluginhub mujtaba3b/gstack-extensions --plugin engComprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Harness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses
Binary reverse engineering, malware analysis, firmware security, and software protection research for authorized security research, CTF competitions, and defensive security
A growing collection of Claude-compatible academic workflow bundles. Covers scientific figures, manuscript writing and polishing, reviewer assessment, citation retrieval, data availability, paper reading, literature search, response letters, paper-to-PPTX conversion, and evidence-grounded Chinese invention patent drafting. Rules are organized as reusable skill folders with explicit workflows and quality checks.
Intelligent draw.io diagramming plugin with AI-powered diagram generation, multi-platform embedding (GitHub, Confluence, Azure DevOps, Notion, Teams, Harness), conditional formatting, live data binding, and MCP server integration for programmatic diagram creation and management.
Persistent file-based planning for AI coding agents. Crash-proof markdown plans (task_plan.md, findings.md, progress.md) that survive context loss and /clear, with an opt-in completion gate and multi-agent shared state. Manus-style. Works with Claude Code, Codex CLI, Cursor, Kiro, OpenCode and 60+ agents via the SKILL.md standard. Includes Arabic, German, Spanish, and Chinese (Simplified and Traditional).