MCP App StoreMCP App Store
from tandem-browser536
Connect to a running Tandem instance (MCP or HTTP) to inspect, interact with, and automate actions in the user's real browser context while preserving safety an
Tandem Browser exposes the capabilities of a running Tandem instance to agents, enabling safe human-AI co-browsing in the user's real browser context. The skill documents MCP and HTTP integration patterns, discovery endpoints, targeting styles (active tab, specific tabId, or session partition), durable handoffs, prompt-injection handling, trust tiers, and recommended workflows for background helper tabs, snapshots, and SPA state mining. It prioritizes safety: prefer snapshots and page-content reads over raw HTML, always verify completion metadata for actions, and use durable handoffs for pauses or approvals.
Use this skill when an agent needs to inspect or interact with tabs that the user already has open, perform actions inside authenticated sessions, coordinate multi-step browser tasks with the user, or run background helper tabs without stealing focus. It's the right choice when the task requires the user's real browser state (cookies, logged-in sessions) rather than a sandboxed headless browser.
Agents that support MCP or can call HTTP APIs and manage tokens — e.g., Claude Code, agent frameworks that speak MCP, and other assistants capable of orchestrating browser-based tasks.
Tandem Browser is a co-browsing skill that lets agents connect to a running Tandem instance via MCP or HTTP to inspect, interact with, and automate actions in the user's real browser context. The SKILL.md is comprehensive and well-structured, covering connection discovery, workspace management, tab targeting, handoffs, sessions, prompt-injection handling, and trust tiers. No bundled scripts were present to test. Minor security deductions: the skill reads local auth tokens from ~/.tandem/api-token and passes them in curl headers (shell var interpolation risk), and instructs agents to execute arbitrary JS in browser context, though both are inherent to the skill's purpose and include appropriate guardrails.
Very thorough and well-documented skill. Strong emphasis on safety: prompt-injection scanning, trust tiers, user-approval modals for dangerous ops, explicit rules against obeying page-embedded instructions. Shell variable interpolation of auth tokens in curl examples is the main security nit. Architecture is exemplary — clean frontmatter, progressive disclosure via discovery routes, clear output contracts per endpoint. No scripts to audit statically.
Development Worktree
Create an isolated git worktree for feature work, auto-run project setup, and verify a clean test baseline before development.
ds-fix — data-science mid-analysis fixer
Orchestrates diagnosis and targeted fixes mid-analysis: diagnose root cause, apply fixes with output-first verification, and update project learnings.
Readwise Reader Document Management
Manage Readwise Reader documents: list, save, search, move, tag, highlight, export and bulk-edit via official and custom CLIs.
Bounty Hunter — Atlas
Persona skill: 'Atlas' — a profit-focused developer persona for discovering, evaluating and executing paid bounties or freelance tasks with ROI-aware workflows.
Full Stack Builder
End-to-end builder that scaffolds, implements, tests, and optionally deploys web and API applications from a natural-language specification.
ezBookkeeping API Tools
Command-line API tools for ezBookkeeping: record and query transactions, retrieve accounts/categories/tags, and fetch exchange rates for self-hosted personal fi
Feishu Voice Sender
Convert MP3s and send them as native Feishu voice messages (playable voice clips) to users or groups.
Claw Bench
Benchmarking skill that guides an agent through a structured suite of capability tests and reporting steps for leaderboard submission.