Bash plugins

Enforce a structured TDD workflow with parallel task execution, isolated git worktrees, and systematic code review. Guides root cause investigation before fixes, verifies tests and linters pass before claiming completion, and dispatches subagents for independent tasks like fixing multiple test failures simultaneously.

Reduce Claude Code token consumption by ~75% using ultra-compressed communication and delegation to compressed-output subagents for code review, editing, and exploration.

Provides a structured engineering workflow: transforms ambiguous ideas into PRDs and GitHub issues, designs module APIs, refactors code via vertical-slice TDD, diagnoses bugs, resolves git conflicts, and produces documentation like ADRs and domain glossaries.

Persists Claude Code context across sessions using a local memory database, enabling the agent to recall past work, decisions, and bugfixes. Includes tools for codebase exploration, architectural analysis, GitHub triage, release automation, and project timeline digests.

Monitor Claude Code sessions with a real-time HUD showing context health, tool activity, agent tracking, and todo progress, configurable via guided setup

Reverse engineer binaries, triage and unpack malware, extract and analyze firmware, perform memory forensics, and dissect network protocols using specialized AI agents and skills for authorized security research, CTFs, and incident response.

Run a comprehensive PR review that delegates to specialized agents for code quality, test coverage, error handling, type design, comment accuracy, and code simplification, producing a categorized issues summary with critical items, suggestions, strengths, and an action plan.

Persist task plans, findings, and progress as markdown files so work survives context loss and /clear in AI coding agents. Manus-style planning workflow with crash-proof session recovery, tamper detection, and multi-language support (Arabic, German, Spanish, Chinese).

Audit CLAUDE.md files across repositories by discovering them with find, evaluating quality against rubrics, generating reports, and applying targeted improvements after approval. Capture learnings from Claude Code sessions to propose concise updates to CLAUDE.md or .claude.local.md files with user approval.

Develop full Claude Code plugins end-to-end: plan and generate agents, commands, skills, hooks, and MCP integrations via guided workflows, then validate structure, naming conventions, and component quality with actionable reviews and fixes.

Run Claude in a continuous self-referential loop that repeatedly feeds the same prompt back to itself, enabling incremental iterative improvement across multiple cycles until a task is completed.

Guides feature development through architectural analysis, codebase exploration, and automated code review, while maintaining project documentation via CLAUDE.md audits, session learnings, and custom skill creation with eval benchmarking.

Research trending topics across Reddit, X, YouTube, TikTok, Instagram, Hacker News, Polymarket, GitHub, and more, with AI-scored results based on engagement metrics like upvotes and likes.

Generate production-ready stateful CLI harnesses for GUI applications from local paths or GitHub repos, implementing Click CLI with REPL/JSON support, pytest unit/E2E tests, and docs. List installed harnesses, refine coverage gaps, run tests to verify functionality, and validate against standards.

Drive creative direction with algorithmic design manifestos, generative p5.js art, brand voice optimization, and portfolio interfaces that convert visitors to clients.

Run a complete academic research pipeline from deep literature review through writing, peer review simulation, revision, and finalization, with citation verification, cross-source synthesis, and AI-usage disclosure generation.

Delegate expert-level code reviews, security audits, penetration tests, QA automation, accessibility compliance checks, performance optimizations, chaos engineering, and compliance validations to specialized sub-agents across codebases, infrastructure, and systems.

Programmatically create, edit, and analyze office documents (DOCX, XLSX, PPTX, PDF) and Google Workspace files using Python scripts, enabling automated report generation, format conversion, and data extraction.

Run PluginEval certification pipeline on Claude plugins or skills to compute quality scores, badges (Platinum/Gold/Silver/Bronze), dimension breakdowns, anti-patterns, and recommendations via static analysis and LLM judging across 10 criteria including triggering, orchestration, and output quality. Compare skills head-to-head or evaluate directories for actionable insights.

Build and manage cloud infrastructure and deployment pipelines with AWS serverless, Docker, Kubernetes, Terraform, and CI/CD workflows, including environment setup, containerization, GitOps, and production deployment strategies.

Execute marketing and growth workflows including A/B testing setup, SEO diagnostics and optimization, analytics tracking audit, programmatic SEO strategy, and automated content/email campaign creation with brand voice analysis.

Write production-grade shell scripts with defensive programming patterns, unit tests via Bats, and static analysis via ShellCheck, enabling robust CI/CD pipelines and cross-platform automation.

Orchestrate swarms of specialized AI agents to automate end-to-end software development: plan features, implement code with Rails/Python/TS patterns, conduct multi-perspective reviews for architecture/security/performance, resolve todos/PR feedback in parallel, run browser/iOS tests, sync Figma designs, generate docs/videos, and ship PRs.

Enforce code quality and security checks after every change, automate git commits and pushes with conventional messages, generate structured task checklists, guide incremental refactoring, and apply systematic debugging—all within Claude Code.

Manage Python projects via structured tracks for features, bugs, refactors: initialize context artifacts like product.md and tech-stack.md, create detailed specs and phased plans, implement tasks with strict TDD workflow using pytest coverage and git commits, monitor status, revert commits, and validate artifacts for consistency.

Autonomous investment banking pitch agent: pulls market data, builds DCF and football-field valuation Excel models, and produces branded pitch deck files on disk — all from a single natural-language request.

Delegate code review, investigation, and complex coding tasks to OpenAI's Codex within Claude Code, with support for adversarial reviews, background job management, and readiness checks.

Automate RevOps and CRM workflows across HubSpot, Stripe, Shopify, Zendesk, SendGrid, Google Calendar, and Outlook Calendar — manage contacts, deals, orders, subscriptions, tickets, campaigns, and events without leaving Claude Code.

Prevents unwanted behaviors in Claude Code by creating, managing, and toggling rules based on conversation patterns, regex, and multi-condition checks for bash commands, file edits, prompts, and stop events.

Sandboxed code execution in 11 languages with automatic session state restore, persistent FTS5 knowledge base for indexed project content retrieval, and hooks that intercept tool calls to conserve context window usage.

Ingest GP valuation packages, run portfolio-company marks through valuation templates, compute NAV and waterfall, and stage LP reporting for IR review in private equity workflows

Automatically ensures headroom hook configuration is up-to-date on session start and before Bash/PowerShell tool calls.

Automates OSS maintenance workflows: changelog generation, conventional commits, PR management, structured documentation, advanced Git operations, and code review handling.

Adds educational insights about implementation choices and codebase patterns to Claude's responses, mimicking the deprecated Explanatory output style.

Manage an AI-supervised issue tracker with persistent task memory, dependency graphs, and compaction recovery, enabling multi-session coding workflows via simple CLI commands.

Manage the full Hugging Face ML lifecycle from a single agent: search and select models, estimate GPU memory, train or fine-tune with TRL/Unsloth, evaluate locally, build and deploy Gradio demos on Spaces, publish datasets and research papers, and run models in-browser with Transformers.js.

Enforces structured troubleshooting and evidence-first delivery through opt-in productivity coaching, with role-based protocols (P7-P10), iterative development gates, and motivational flavors. Automatically activates on failure or passivity cues to push developers toward thorough, verifiable completions.

Debug GitHub Actions failures by analyzing logs, identifying breaking commits, and finding fix PRs. Clone conversations to branch experiments or halve token usage. Generate HANDOFF.md for agent handoffs and review chats to improve CLAUDE.md. Fetch Reddit content when WebFetch is blocked.

Prompts developers to make meaningful code contributions at decision points during coding sessions, creating an interactive learning environment that mimics a guided tutorial style, initialized via shell scripts.

Automates equity research workflows: reads earnings transcripts and filings, updates financial models with new data, generates professional post-earnings reports with variance analysis, and drafts morning meeting notes — all without live Excel, using Python/openpyxl for .xlsx output.

Automate KYC/AML screening of investor onboarding documents: parse identity, ownership, source of funds; apply risk rules; check sanctions/PEP lists; generate escalation packets for compliance sign-off.

Trace execution flows, analyze blast radius, and search codebases using a knowledge graph. Index Git repos, trace bugs through function callers/callees, assess code change impact, review pull requests with risk detection, and safely refactor with dry-run previews.

Run an autonomous engineering pipeline from ideation through PR: brainstorm requirements, plan, implement, simplify, review, test, commit, push, open PR, and watch CI until green. Also debug errors, run QA passes, generate release notes and marketing copy, and compound team knowledge with structured solution docs.

Automate general ledger to subledger reconciliation: detect breaks, trace root cause to originating journal entries, classify break causes, and route exception reports for sign-off. Also audits spreadsheets for financial model errors and integrity.

Controls iMessage channel access for Claude Code — approve/deny pairings, manage sender allowlists, set DM and group policies, and verify setup including Full Disk Access and current policy status.

Persist memory across AI coding agent sessions by capturing tool usage and insights, compressing via LLM, and injecting relevant past context into future interactions. Recall session history, search observations, and forget specific data via natural language commands.

Build and audit financial models (DCF, LBO, comps, 3-statement) in Excel, generate competitive landscape and pitch decks in PowerPoint, and access financial data from FactSet, PitchBook, Morningstar, and other sources.

Turns any Claude conversation into a persistent, self-organizing Obsidian wiki vault with hybrid retrieval, methodology-based filing (LYT/PARA/Zettelkasten/Generic), automated research loops, canvas management, pre-commit auditing, and health checks — enabling compounding knowledge that survives across sessions and projects.

Enable Claude Code to autonomously curate its memory, promote proven patterns to permanent CLAUDE.md rules, and extract reusable skills from recurring solutions, improving code quality and automation over time.

Build institutional-quality financial models (DCF, LBO, three-statement, and trading comparables) directly in Excel from a ticker and assumption set, with automated formula linkages, integrity audits, sensitivity tables exports as .xlsx files.

Audit and optimize paid advertising campaigns across Google, Meta, YouTube, LinkedIn, TikTok, Microsoft, Apple, and Amazon Ads with 250+ automated checks, weighted scoring, AI creative generation, and deep-dive attribution analysis.

Query biomedical literature and preprints (PubMed, biorxiv), search clinical trials, explore drug-target associations (Open Targets, ChEMBL), and run preclinical analyses (RNA-seq, single-cell QC, scvi-tools) directly from the coding environment

Autonomously optimize code files by measurable metrics through iterative experiments: set up target file, eval command, and loop intervals (10min-monthly); AI edits code, commits to git branches, evaluates with Python, keeps improvements. Resume, run manually, or check dashboard status.

Automate month-end close tasks including accruals, roll-forwards, and variance commentary, and assemble a close package for controller sign-off.

Audit LP capital-account statements against the fund NAV pack before distribution — ties out balances, allocations, and fees, flags discrepancies, and checks spreadsheet formulas for errors, hardcodes, and circular references.

Manage AI-driven development workflows with hierarchical task trees, dependency graphs, automated subtask expansion, PRD-to-task parsing, status tracking, and intelligent task orchestration via natural language commands.

Automatically recall relevant memories and retain conversation transcripts across Claude Code sessions, enabling persistent context and long-term knowledge retention without manual management.

Spawn parallel AI subagents in isolated git worktrees to compete on tasks like code optimization, refactoring, test writing, or bug fixing. Evaluate results using pytest metrics or LLM judging on git diffs, rank agents, and merge the top performer into your base branch.

Automate web tasks with natural language: drives a local Playwright browser, captures screenshots and action logs, visually self-verifies results. Also generates reusable Python CLI tools from the workflows.

Full-cycle development workflow for solo developers: project scaffolding, multi-agent collaboration (Cursor PM & Claude Code), CI/CD automation, auth/payments setup, code reviews, release management, and cross-session memory.

Run CodeQL and Semgrep to scan multi-language codebases (Python, JavaScript/TS, Go, Java, C#, Ruby, Rust) for security vulnerabilities via taint tracking and pattern matching. Parse, deduplicate, and aggregate SARIF outputs from scans, then integrate findings into CI/CD pipelines using GitHub Actions or bash scripts.

Spawn AI agents, create workspaces, manage tasks, and schedule automations using Superset directly from Claude Code. Edit automation prompts in your preferred editor, view recent logs, and trigger one-off runs with simple commands.

Mark up and refine AI-generated plans interactively in a UI, annotate markdown files, messages, and git changes for review, share for team collaboration, browse plan archives, and automate workflows with plan mode hooks.

Enhances AI-assisted development for Chinese-speaking teams by providing structured workflows for planning, coding, testing, debugging, code review, and Git management, all adapted to Chinese conventions and domestic platforms.

Implement Trail of Bits handbook security testing workflows: fuzz Rust, Python, C/C++, Ruby code with AFL++, libFuzzer, cargo-fuzz, Atheris; instrument AddressSanitizer; run static analysis via Semgrep, CodeQL; generate coverage reports, dictionaries, and bypass obstacles for vulnerability detection.

Run 10 AI agents to fully automate Obsidian vault management: triage Gmail/Hey emails and inbox notes, extract deadlines from Google Calendar, transcribe audio into structured notes, audit and defragment vault structure, generate weekly agendas, evolve knowledge graph, and handle multilingual inputs.

Run institutional-grade stock analysis for A-shares, HK, and US equities: 22-dimension data, 65-investor panel voting, 17 valuation methods (DCF, LBO, comps), fraud detection, and Bloomberg-style HTML reports. Supports portfolio drift analysis, earnings preview, catalyst calendars, and investment thesis tracking.

Run system commands, manage processes, and access files beyond the workspace from within Claude, including structured documents like PDFs, spreadsheets, and images, with persistent shell state across conversations

Apply Maoist dialectical reasoning and strategic frameworks to software development: prioritize tasks via contradiction analysis, resolve trade-offs, investigate unknowns, self-criticize completed work, and bootstrap projects from zero resources using phased warfare tactics.

Crystallize vague project requirements into executable Seed specifications through Socratic interviews, then run, evaluate, and iteratively refine them with three-stage verification, drift detection, and evolutionary loops.

Build multi-language code graphs to map call graphs, attack surfaces, blast radius, taint propagation, privilege boundaries, and complexity hotspots for security audits. Visualize architecture with Mermaid diagrams, compare snapshots across git commits for evolution analysis, triage mutation testing survivors, generate crypto test vectors, diagram protocols, and project SARIF findings onto graphs.

Improve Claude's initial output quality by automatically injecting context, clarifying vague prompts with research-based questions, and applying targeted nudges for plan readability, workflow routing, background execution, and subagent routing.

Runs Mooncake pre-PR validation locally to catch typos, formatting, build and test failures before pushing, using a single shell script that mirrors the CI pipeline's reproducible checks.

Automate long-form webnovel creation: initialize projects interactively with genre/characters/worldbuilding/outlines, generate beat sheets/chapters (2000+ words), extract entities/relationships to SQLite indexes, visualize status/entity graphs in read-only dashboard, recover interrupted workflows, and validate chapters via agents for inconsistencies, pacing, OOC, reader pull, and quality reports.

Build and orchestrate advanced Claude Code agentic workflows by creating meta-prompts, subagents, hooks, MCP servers, slash commands, and skills; execute hierarchical plans, run autonomous coding loops, apply expert debugging and productivity frameworks like 5 Whys or Eisenhower Matrix, and audit components for compliance and quality.

Audit and fix token waste across Claude Code and Codex sessions. Recover 5-25% of context budget through config cleanup and compaction, monitor usage trends via a live dashboard, and get real-time coaching on efficient prompt design.

Draw cryptographically random Tarot card spreads via Bash and Python to interpret context, action, obstacles, and outcomes for guiding decisions on vague or ambiguous coding prompts and planning. Invoke for entropy-driven insights when feeling lucky or prompts lack specificity.

Build Claude Code plugins by replicating demonstrated skills, legacy commands, and MCP server structures. Copy SKILL.md layouts, frontmatter options like name-description-argument-hint-allowed-tools-model, argument parsing, tool integrations (Read, Glob, Grep, Bash), reference templates, and content best practices for quick extension development.

Generate complete Ansible playbooks, roles, inventories, and configurations to automate server provisioning, app deployments, service setups, and idempotent DevOps workflows with production-ready YAML, best practices, security hardening, and built-in documentation.

Generate production-ready Helm 3 charts for Kubernetes apps from natural language descriptions. Produces Chart.yaml, values.yaml with multi-environment overrides, Go templates, helpers, health probes, security contexts, and applies best practices for scalability, security, and documentation.

Read, send, and manage Gmail messages; create and edit Google Docs, Sheets, and Slides; access and organize Drive files; manage Calendar events, Tasks, and Contacts; orchestrate Chat messages, Apps Script execution, and Custom Search via MCP tools from the command line or uvx.

Streamline end-to-end Obsidian plugin development and vault management: scaffold projects with TypeScript setups, implement UI views/events/data handling, optimize performance/security, establish local dev loops/CI/CD/release pipelines, migrate content, and troubleshoot errors using 24 specialized skills.

Manage Google Ads and Meta Ads campaigns, run SEO audits, optimize content for AI search engines (GEO), and generate schema markup — all from Claude. Includes keyword research, broken link scanning, landing page scoring, and content calendar planning powered by Google Search Console and advertising APIs.

Generate professional Excel financial models including DCF valuations with FCF projections, WACC, sensitivities; LBO analyses with debt schedules, IRR/MOIC; budget vs actual variance reports with waterfalls; and dynamic pivot tables via natural language prompts and auto-invoked skills.

Run end-to-end YouTube content strategy workflows: research competitors via channel scraping and analysis, generate tiered video ideas with validation, produce structured briefs and detailed outlines including demo prep, craft CTR-optimized titles and thumbnail concepts.

Audit codebases with a security agent that scans for vulnerabilities like SQL injection, XSS, CSRF, auth flaws, insecure dependencies, and secrets; generates severity-rated reports including file locations, explanations, compliance checks, and code fixes with examples.

Automatically discover and hierarchically load AGENTS.md files across project directories into Claude's agent context, merging instructions with conflict detection and caching for specialized behaviors without manual setup. Sync all agent contexts into CLAUDE.md under organized sections with backups and summaries.

Set up Ollama for local AI model inference on macOS, Linux, or Docker with automated installation, hardware-optimized model selection, GPU configuration, verification, model pulls, API testing, and client integration via Python, Node.js, or REST for zero-cost, privacy-first LLM workflows.

导入已有中文网络小说（半成品或完本）并自动解析为标准项目目录结构，支持长/短篇路由分流，打通从导入→分析→去AI化→封面设计→趋势研究→继续写作的完整创作管线，配合Agent协作与路由系统实现网文写作的逆向工程与正向产出闭环。

Create and validate custom Semgrep rules for detecting security vulnerabilities, bugs, code patterns, and standards using test-first methodology, conversation context for patterns and languages, plus taint mode support.

Monitor usage limits, reset times, and cost across multiple AI CLIs (Claude, Codex, Gemini, z.ai) from a unified terminal dashboard, with automatic recommendations for the CLI with most available capacity

Extends Claude Code with spec-driven development workflows: autonomous long-running tasks, multi-agent deep research, structured feature planning with EARS requirements and constitution-based specs, GitHub issue fixing and PR review, skill benchmarking and creation, image generation, and session learning capture.

Automatically reviews and fixes Claude Code skills through iterative refinement until they meet quality standards, with commands to cancel loops and preserve changes.

Run AI-powered code reviews on uncommitted changes, branch diffs, or specific commits using external LLM CLIs (OpenAI Codex, Google Gemini), with an optional MCP server for direct tool access and collaborative reasoning.

Automatically discovers, extracts, and adapts skills from 229 plugins to tackle your current task without manual selection. Analyzes commands, agents, skills, and scripts across the repository, then synthesizes and applies tailored patterns. Invoke with /skill-adapter or enable auto-activation on relevant requests.

Scaffold production-grade Claude Code plugins with marketplace integration, validate structure and schemas, audit for security vulnerabilities and best practices, and automate semantic version bumps across manifests and catalogs using auto-invoked skills and interactive commands.