uv plugins

Run continuous self-referential AI loops that iteratively prompt Claude with prior outputs, git changes, and file references until task completion or max iterations reached. Start loops with /ralph-loop, cancel with /cancel-ralph, and track progress via local files.

Build scalable production Python backends and APIs with Django 5.x async views, FastAPI microservices, Celery tasks, SQLAlchemy/Pydantic data handling, pytest testing strategies, and architecture optimizations using uv/ruff for modern 3.12+ codebases.

Enforce rigorous QA and testing workflows in Claude Code sessions: drive TDD for features and fixes, debug via four-phase root cause analysis, automate browsers with Playwright/Puppeteer best practices, plan A/B experiments with gates, apply code review checklists, build reliable E2E suites, and triage pytest failures systematically.

Crystallize vague requirements into precise YAML Seed specs via Socratic interviews and ambiguity scoring, then execute workflows, run 3-stage evaluations (mechanical/semantic/consensus), detect goal drift, and iteratively self-improve through evolutionary loops with Git integration.

Automate iterative reviews and fixes for Claude Code skills using a reviewer agent in loops until quality standards are met. Target skills by path or name with max iterations, cancel active sessions by ID while preserving changes, and run custom Python hooks at session end.

Generate, edit, and inpaint images via GPT Image 2 CLI skill, using a reference prompt gallery to match styles for UI mockups, diagrams, posters, research figures, anime, and Chinese typography workflows.

Automate Google Workspace operations like checking Gmail, searching Drive files, scheduling Calendar events, editing Docs and Sheets via a local OAuth-secured MCP server and CLI tools with 114 routed commands for seamless workflow integration.

Receive outbound phone calls from Claude for real-time voice input, progress reports, and task discussions when text falls short. Twilio handles calls with OpenAI-powered speech-to-text transcription and text-to-speech synthesis, plus session-end automation via Python hooks.

Engage in natural voice conversations with Claude Code using local Whisper speech-to-text and Kokoro text-to-speech. Auto-activate via mentions or commands, play background music, clone custom voices from clips, connect remotely to iOS/web clients, install dependencies, and monitor service status.

Automatically format Markdown files using Prettier whenever a Claude Code session ends via Stop hook. Exclude specific organizations or paths through configuration to target only relevant files and maintain workflow productivity without manual intervention.

Automate iterative code review loops: supply a task, Claude implements it, Codex performs independent CLI review, Claude refines based on feedback until complete. Cancel active loops and hook custom actions on session end.

Initialize and run an autonomous AI copywriter agent that generates SaaS marketing content from PRD stories: learns your writing voice, researches competitors, drafts and iterates copy, runs npm tests, commits via git. Monitor pipeline status with DB, progress, and blockers; cancel loops as needed.

Hear a short spoken summary of Claude's response using macOS 'say' command. Automatically runs at the end of each Claude Code session via a Python hook, enabling quick audio review of outputs for better accessibility and hands-free workflow.

Drive spec-driven development via SpecKit workflow: define project constitutions, generate feature specs/plans/tasks, execute autonomously with multi-agent coordination, git branching/commits, QA verification, and progress tracking across phases.

Automatically evaluates Claude Code sessions on stop events to decide if continuation is needed, using Claude-judged decision making via Python hooks, preventing premature workflow termination and enabling seamless extended sessions.

Track carbon footprint of Claude Code sessions using SQLite database queries for reports on daily and yearly CO2 emissions, equivalences to real-world impacts, top sessions by emissions and costs, and per-project breakdowns. Generate shareable PNG report cards exported to ~/code/claude-carbon/exports/. Runs automated hooks at session end.

Scaffold Nx workspaces and generate AWS cloud-native apps, including TypeScript/Python backends, React frontends, tRPC APIs, Lambda functions, and infrastructure via CDK or Terraform.

Configure OpenAI or OpenRouter for AI-powered code reviews on uncommitted changes. Trigger automatic standard reviews, adversarial probes of design assumptions and trade-offs with severity ratings, or rescue tasks delegated to Codex for fixes and analysis in the background.

Launch parallel multi-agent deep research on topics using Claude and Codex, enabling cross-pollination refinement for structured synthesis reports. Cancel active sessions via CLI and automate post-session hooks.

Send and receive iMessages for real-time notifications and user input during Claude Code sessions. Toggle text mode with /textme to enable updates on task completion, decisions, or clarifications via Sendblue SMS/iMessage, allowing remote interaction without terminal monitoring.

Manage Pinecone vector indexes and build RAG/semantic search apps directly in Claude: create/upsert/query via MCP server and skills, quickstart Python projects with /quickstart, explore indexes with /query, ingest documents, and handle full-text/hybrid search.

Build end-to-end bridgic-amphibious Python AI agent projects: initialize OpenAI/vLLM providers, create dual-mode LLM agents with human-in-loop fallbacks, automate browser tasks like scraping and UI ops, scaffold via CLI orchestration, and generate runnable codebases from task descriptions.

Automate Jira issue management via CLI scripts triggered by URLs or keys: search/create/update/transition/comment/log work/attachments/sprints/boards. Convert Markdown to Jira wiki markup with bug/feature templates and syntax validation. Run custom Python hooks before prompt processing for further automation.

Build annotated collections of research papers by retrieving PDFs from arXiv, DOIs, titles, or sci-hub; extract structured notes, claims, concepts, justifications from them; cross-reference citations, identify new leads, tag, audit directories, and integrate into propstore knowledge stores via CLI workflows.

Run autonomous Claude-Codex loops to plan features: Claude drafts PLAN.md, Codex delivers adversarial critiques with pushback, auto-revises over rounds until LGTM or maxed. Audit git diffs via Codex for findings and fix patches in reviews/. Clean up stuck states, check status, cancel cleanly.

Route LLM tasks like research, code, analysis, and generation to cheapest capable models across 20+ providers via auto-classification by type and complexity, monitor subscriptions, track savings with dashboards and alerts to Slack/Discord, and automate llm-router releases to PyPI/GitHub.

Activate a Codex-powered code review gate in your git repo before a Claude Code session. On session exit, it automatically triggers a CLI review with configurable focus areas and max cycles. Cancel gates anytime to deactivate.

Operate Armorer-managed applications via unified CLI to install, uninstall, run, configure, inspect state, monitor jobs, queues, logs, and verify runtime health, replacing ad-hoc shell commands.

Run iterative Ralph development loops in git repos: Claude codes on tasks with max iterations, gated by Codex CLI reviews before completion. Start loops with task prompts and flags, cancel active sessions, view status and help.

Run Claude in continuous self-referential loops that iteratively execute development tasks by re-prompting with preserved file changes and git history until completion or max iterations. Cancel active loops and explain the process.

Orchestrate multi-agent AI pipelines via CLI to autonomously study codebases, strategize improvements, build PRs, review changes, evaluate outcomes, and apply successful evolutions. Analyze quality, track status, and execute custom agent DAGs for software evolution workflows.

Formalize scientific papers, textbooks, or knowledge sources into structured Gaia knowledge packages via a six-pass process: extract propositions, connect reasoning, assign priors, compile DSL to IR, then generate Obsidian vaults, Mermaid graphs, READMEs, and publish to GitHub.

Develop, test, debug, and deploy Fastly edge compute apps and CDN services end-to-end: run Rust/Go WASM locally with Viceroy, manage services/backends/VCL/security via CLI, lint/format VCL, transpile XVCL, and audit NGWAF rules for production readiness.

Automatically title Claude Code sessions and trigger sound notifications at session end using Python hooks run via uv on stop events. This unlocks seamless workflow feedback and organization without manual intervention, enhancing productivity in coding sessions with auditory cues and smart naming.

Automatically capture and send telemetry from Claude Code sessions to Grafana Sigil on session stop, enabling observability tracking of coding workflows via a Python hook executed with uv.

Clone untrusted Python dependencies from GitHub, decompose them into verifiable sub-packages via test-driven evaluation, generate focused pytest unit tests, rewrite imports, and iteratively implement secure from-scratch replacements to mitigate supply chain attacks.

Synthesize Zotero collections of academic papers into theoretical maps, thematic clusters, and identified debates for deep literature analysis before writing research papers or reviews.

Analyze survey microdata with weighted pandas DataFrames to compute Gini coefficients, poverty rates, quantiles, and inequality metrics. Impute missing values using ML methods like random forest and XGBoost from donor data. Calibrate weights to population targets with L0 regularization. Enhance datasets like CPS ASEC and run PolicyEngine microsimulations for tax-benefit policy impacts across populations.

Run PRD-driven iterative coding loops in Claude Code: interview users to generate scoped PRDs, decompose into git-tracked stories, orchestrate multi-agent parallel execution, monitor progress via state files and commits, start/stop/plan/cancel with slash commands.

Execute autonomous BMAD-method workflows for SDLC automation, such as generating product briefs and PRDs via ralph-loop with templated prompts auto-populated from project context. Track phased status as DONE/READY/BLOCKED tables, view plugin docs, cancel active loops, and run custom hooks at session end.

Automate end-to-end maintenance of Python/ML open-source projects on GitHub: triage and analyze issues/PRs/discussions, resolve PR conflicts with attributed fixes, conduct multi-agent code reviews for quality/security/performance, prepare release artifacts/changelogs/migration guides, and optimize GitHub Actions CI/CD pipelines.

Orchestrate autonomous Claude-based research agents to investigate hypotheses, spawn experiment runners that execute Python scripts and analyze outputs with LLM judges, generate interactive Quarto HTML reports, send supervisor notifications, perform fresh-eyes reviews, and run batch evaluations.

Download videos from 1800+ platforms like YouTube, TikTok, and Twitter, then auto-generate MP4 files, MP3 audio extracts, VTT subtitles, TXT transcripts, and Markdown AI summaries. Everything saves directly to your downloads folder for easy access and reuse in projects.

Automate full dev lifecycle: summarize branches and uncommitted work, generate conventional commits, create/triage GitHub PRs, polish code per guidelines, safely upgrade deps with pnpm/uv, extract action items from transcripts/Slack into Linear tickets, and produce handoff docs.

Initialize and manage a project-scoped knowledge vault: ingest Markdown clippings, URLs, text, or batch academic papers from research servers; compile into wiki summaries with cross-references; query for answers citing sources via wikilinks; audit, lint, and fix content quality; browse vault in Obsidian.

Run Dr. Ralph's iterative AI diagnostic loops for development tasks: input symptoms (code issues), AI interviews, web researches, generates differential diagnoses, treatment plans, SOAP reports with notes, looping until completion; cancel anytime.

Automate web monitoring with scouts tracking API changes, changelogs, competitor news, funding, and product updates every 12 hours or daily. Conduct deep agent-based research synthesizing multi-source web info. Automate browser tasks like form filling, data extraction, and multi-step navigation via Yutori Navigator.

Develop, test, debug, review, and migrate Keboola Python components for data pipelines—including extractors, writers, apps—with AI skills for config schemas (conditional fields, UI), code quality (Ruff, architecture), local datadir/pytest/VCR testing, uv/pyproject.toml upgrades, Docker builds, and platform context via MCP/Datadog.

Deploy HIPAA-compliant guardrails for FHIR AI agents handling US Core R4/R6 data: redact PHI on reads, enforce step-up auth for writes, log immutable audits, isolate tenants, evaluate/correct data quality with Curatr, ingest EHR from Epic/Cerner/Athena via Fasten Connect, export portable records via HealthEx to local FHIR stores.

Run Claude in continuous self-referential loops for iterative task development, repeating the same prompt until a file-based completion promise is met or max iterations reached, with git integration for versioning. Cancel active loops, explain the technique, and automate end-of-session hooks.

Deploy Ralph, an autonomous agent, to automate iterative development by creating PRDs, preparing tasks, executing runs, and monitoring status across multiple parallel PRDs using slash commands like /prd, /prep, /run, /status directly in Claude Code sessions.

Debug AI chains by inspecting and managing LangSmith traces, runs, datasets, and prompts via CLI; analyze token costs and follow cache-first workflows for efficient observability and evaluations.

Generate domain-specialized images like banners, ads, YouTube thumbnails, illustrations, logos, product photos, portraits, and web assets via Google Gemini. Uses auto-built prompts, keyword triggers, format/aspect ratio detection, and reference image editing for quick design workflows.

Extract text and structured data from single or batch PDFs to Markdown or TXT via CLI, using 9 backends with automatic fallback and OCR for scanned documents.

Execute multi-step AI agent workflows on Linear tickets via natural language CLI commands, chaining headless ai-cli runs with branching logic for triage, implementation, status checks, and inspections. Automate custom Python hooks at Claude Code session end using uv.

Index multimodal directories, explore/search files, extract text/images/keyframes from PDFs/videos, and process visuals via natural language with Orion VLM agent for OCR, object detection, summarization, and multi-modal generation.

Orchestrate end-to-end Ansible workflows: scaffold playbooks and roles, generate idempotent code with best practices, validate syntax and lint with ansible-lint, review for security/performance, debug failures, and automate Proxmox VM/cluster management via specialized agents and commands.

Bootstrap Claude Code with 17 specialized agents, skills, and hooks to audit/evolve .claude/ configs, engineer/refactor Python code via TDD, profile/optimize ML workloads, generate docs/tests, design systems, diagnose issues, and manage workflows professionally.

Automate end-to-end Agisoft Metashape photogrammetry workflows via MCP server: import photos/GPS/masks, align cameras with drift detection/QA, generate dense clouds/meshes/textures/DEMs/orthos, diagnose alignment/texture issues, remove artifacts, export tiles for Blender/Unity/Godot, integrate USGS LiDAR.

Run fully autonomous AI coding sessions that generate step-by-step plans, execute tasks with self-critique and best practices, enforce file policies to prevent clutter, block risky commands via safety guards, manage tmux for isolated testing, and export refined plans.

Run specialized AI agent skills to brainstorm designs, conduct multi-agent research and task decomposition, implement verified code with cross-model reviews, manage Kubernetes Tilt dev and agent sandboxes, handle Python uv/ruff workflows, perform security audits and git ops, plus consolidate session knowledge via dreaming cycles.

Automate end-to-end Linear workflows from CLI: triage issues into prioritized tickets, create git branches with ticket IDs, commit/push changes, and open linked GitHub PRs; manage projects, milestones, documents, and cross-repo coordination.

Evaluate and iteratively improve AI agent skills: generate eval.yaml configs and test cases from SKILL.md, run judge-scored evals with MLflow logging, auto-edit fixes for failures, detect regressions, and interactively review results.

Manage Bauplan data lakehouses end-to-end: safely ingest S3 data via Write-Audit-Publish with quality gates and atomic merges, explore schemas/tables/profiles using Python SDK, create SQL/Python transformation pipelines, generate Polars-based quality expectations, assess data fitness for business questions, and debug failed jobs with pinned snapshots and Git fixes.

Automate ASDA grocery shopping entirely from the terminal by searching products, managing your basket with add/remove/view actions, authenticating your account, booking delivery slots, and placing orders.

Run endless autonomous optimization loops on code targets like LLM training loss, test speed, bundle size, or build time: plugin edits files, commits via git, executes short experiments or benchmarks, measures metrics, keeps improvements, reverts failures, and iterates until manually stopped.

Perform thorough reviews of Python tests in code or projects using a standard checklist that evaluates isolation, mocks, execution time, flakiness, and naming clarity to uphold high testing standards.

Evaluate Claude Code harness engineering quality using quick checklist scoring, standard static/dynamic analysis, or full multi-agent reviews across safety, completeness, and design. Track score trends over time with comparisons, generate bilingual reports, grades, badges, and prioritized improvement roadmaps in minutes.

Summon Python specialists to scaffold production Django/FastAPI projects with uv/Docker/PostgreSQL, enforce Mypy/PEP8/security reviews, audit codebases for multi-agent parallelization, implement Celery tasks/WebSockets, and generate pytest strategies—all using 2025 patterns and official docs.

Build TypeDB-powered notebooks for scientific knowledge: ingest literature into a knowledge graph, retrieve via semantic search and dynamic TypeQL queries, analyze with agentic 6-phase curation workflows, index bioskills for composition, and automate browser tasks.

Extract key learnings from Claude Code conversation transcripts and curate them into a personal skills library at ~/.claude/skills/SKILL.md. The agent identifies reusable knowledge, edits existing files for quality, limits to 15 top skills with full rebuilds on overflow, and triggers a custom Python hook via uv at session end for further automation.

Map your repo structure to detect naming clashes and doc gaps, track cross-session goals and learnings, build evolving narratives from git history, and auto-generate READMEs—persisting tribal knowledge for fast code exploration.

Run four-layer tests on Claude Code plugin skills to validate structure, check trigger accuracy, simulate multi-turn sessions, and compare skill outputs using Python scripts and Bash commands.

Reformat markitdown-generated Excel Markdown files into clean, readable multi-sheet documents via analyze-transform-merge pipeline. Convert .xlsx files or directories directly to validated reformed.md outputs, cleaning NaNs, unnamed columns, empty rows/columns, and ensuring table integrity with parallel sheet processing.

Enable voice-first interaction with Claude Code by running a local TTS server on localhost:5001. Switch modes for deep 70-word analyses, 1-sentence acknowledgments, behavioral questions, or 25-word code explanations with interruption handling. Manage server start/stop, voices via JSON, and session-end hooks.

Track Claude Code session usage—tokens, costs, skills, MCP, subagents—by sending data to a self-hosted dashboard. Check dashboard status to confirm tracking is active and automate custom Python scripts at session end via hooks.

Run a local MCP server to index Markdown vaults for full-text and semantic search via embeddings, analyze link graphs, retrieve context, and safely edit notes with git integration—no API keys required.

Set up a continuous self-referencing AI loop in Claude for interactive iterative development, implementing the Ralph Wiggum technique to repeat the same prompt in a while-true cycle until the task completes. Runs a custom Python hook handler via uv on Claude Code session stop for end-of-session automation.

Flag corrections to train Claude on your coding style across sessions, view evolving playbooks and user profiles, force immediate learning extractions, manage Reflexio backend/dashboard lifecycle, and monitor progress via local dashboard.

Automate full Jira task lifecycle in your IDE: create/start issues with git branches/worktrees, generate approach docs/implementation/tests/reviews, create PRs or local merges, run lints/tests, post results/updates/reports to Jira, manage local dashboard, and cleanup branches via commands, skills, and agents.

Launch surveys, preference comparisons, ratings, and rankings to collect real human opinions directly within Claude Code, enabling quick feedback on code, designs, or ideas via Datapoint AI integration.

Intercept bash tool calls before execution and route them to optimal alternatives, such as gh CLI instead of WebFetch for GitHub PRs, while validating plugin installation, hooks format, and route patterns via manifest checks and pytest in a monorepo setup.

Securely install, scan, audit, and monitor packages across 14 ecosystems (npm, yarn, pnpm, bun, pip, uv, poetry, cargo, go, gem, composer, nuget, hex, pub) with CVE detection via OSV.dev, auto-fixes, severity triage, dependency audits, and continuous email alerts using INFYNON CLI and agent.

Automate code quality enforcement and security guards in Claude Code sessions by triggering Python scripts via uv on events for linting, prose checks, atomic writes, bash overflow detection, image blocking, and compaction suggestions. Runs bash commands and writes files safely.

Automate document processing workflows with Nutrient DWS API via Python client and uv CLI scripts. Convert, merge, split PDFs; run OCR; extract text, tables, key-value pairs; apply watermarks, redactions, signatures; optimize, protect, reorder pages; chain into custom multi-step pipelines.

Run adaptive autonomous SDLC workflows that orchestrate agent teams to implement Python features via enforced TDD/BDD cycles with pytest-bdd scaffolding, git worktree isolation for parallel tasks, Beads CLI for dependency-tracked issue management, ruff/mypy/pytest verification pipelines, documentation updates, PR creation, and automated merges.

Persist decisions, specs, and architectural constraints across Claude Code sessions as a context companion. Bootstrap canonical OpenSpec by interactively setting product details, discovering codebase modules via Bash, and mining historical sessions into a module registry. Initialize sessions suggesting recent git changes, load or reconcile prior specs, with hooks automating on edits and session events.

Automatically scan Claude responses for hallucination patterns, speculative language, and unverified claims before delivery, ensuring reliable AI outputs in your coding sessions. Triggers as a hook on session stop events via Python script for seamless workflow protection.

Capture and sync task learnings from Claude Code sessions to Mem0 and Obsidian at session end. Review content before saving, with option to generate reusable agent instructions for ongoing knowledge retention across workflows.

Enforce strict code reviews in git repos using OpenAI Codex as an independent auditor: start sessions with issue ledgers, review git diffs for findings, address or close them explicitly, and finalize only when no blocking issues remain, preventing premature completion claims.

Scaffold production-grade Python packages with best-practice configs for pyproject.toml, Ruff/mypy linting, pytest testing, GitHub Actions CI/CD, MkDocs docs, wheels/sdists packaging, CLI/API design, versioning, and security hardening—or audit repos for violations with structured reports, fixes, and health scores.

Run a persistent 3-screen TUI side panel in Claude Code with free-form main area, auto-updating status dashboard, and ambient screensavers curated by background AI. Receive Distillery knowledge reminders on session start/stop and access local file system via stdio MCP server powered by uv.

Develop STM32 firmware with AI agents that generate HAL peripheral init code, debug HardFaults and bus issues, search full documentation, optimize low-power modes, configure bootloaders and security, analyze safety compliance, and review hardware designs.

Run continuous self-referential Claude loops for iterative development: repeatedly prompt with the same input, incorporating file updates and git changes until task completion or max iterations reached. Start via /ralph-loop with custom prompt and options like --max-iterations, cancel with /cancel-ralph.

Run persistent marathon development sessions on Linear issues across multiple Claude Code sessions, with automatic state management, progress tracking, task creation/updates, and querying of cycles/roadmaps via Linear integration.

Initialize a project and run an autonomous AI agent that generates SaaS marketing content from PRD stories: analyzes your writing voice, researches competitors, drafts copy, self-critiques and iterates until polished, runs tests, commits to git, with pipeline monitoring and cancellation.