MCP App StoreMCP App Store
from humanize50
Generate or rewrite Chinese communication copy to sound more natural and human, with a scorer-driven optimization loop and visible reports.
Humanize is a focused optimization pipeline that takes a Chinese communication task or draft and iteratively improves it to sound more natural and human. It bootstraps a local runtime, generates baseline and challenger drafts, scores candidates with a local reranker, and produces a visible report with scores and decisions.
Use when you need to improve Chinese copy for customer messages, social media posts, emails, WeChat replies, or other short-form communications and want a defensible, scored rewrite rather than a single manual edit. Trigger examples: "make this message sound more human in Chinese", "rewrite my WeChat reply to sound natural".
humanize.py) for running end-to-end optimization, plus bootstrap and scoring scripts as documented in SKILL.md.python3 humanize.py --text "..." --output-root ./runs), and non-negotiable behavior for relaying final outputs and handling timeouts. It documents inputs, default assumptions, and how to interpret the generated reports.Intended for local execution with Python (requires python3) but callable by agents that can run shell commands; suitable for CoPaw/OpenClaw, Claude Code, or any agent that can run the documented CLI.
Humanize is a Chinese copy optimization skill that generates or rewrites text to sound more natural and human. It uses a local scorer model (BAAI/bge-reranker-v2-m3) with a multi-round challenger/baseline comparison loop. Scripts are well-written but heavily coupled — most fail due to cross-imports (runtime_common) not being on sys.path when run from a temp directory. The CoPaw runtime dependency (qwenpaw/copaw) is a hard requirement for generation. Only 2 of 18 scripts ran cleanly (runtime_common, strategy_state as pure imports); the rest failed on missing modules or missing arguments. The skill requires significant setup: CoPaw runtime, torch, a downloaded model, and python3-venv.
torchtransformerspython3-venvqwenpaw/copaw (CoPaw runtime)Well-engineered skill for a specific niche (Chinese copy humanization). The cross-import design means scripts can't run from arbitrary directories. The model download lacks checksum verification — minor security concern for a model hosting platform (HuggingFace). No exfiltration, no hardcoded credentials, no destructive operations. The CoPaw dependency is clearly documented in SKILL.md.
Readwise Reader Document Management
Manage Readwise Reader documents: list, save, search, move, tag, highlight, export and bulk-edit via official and custom CLIs.
SHIP Score Interview
Run a structured SHIP interview to score and prioritise product features or bugs (Income + Strategy - Perspiration). Guides the agent to reach 90-95% confidence
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
Tmux Bridge
Give an agent controlled access to a local tmux terminal bridge so it can run shell commands, capture output, and manage sessions on the user's machine.
Nit (Nitter CLI)
Terminal client to browse tweets, view profiles, and search posts via Nitter instances without opening a browser.
hn — Hacker News CLI
Terminal-based Hacker News client with a CLI and optional interactive TUI for browsing top/new/best/ask/show/jobs stories, viewing threads with nested comments,
Contagious — STEPPS Virality Framework
Apply the STEPPS (Social Currency, Triggers, Emotion, Public, Practical Value, Stories) framework to engineer word-of-mouth and make content, products, or campa
Lokalise Migration — Deep Dive
Step-by-step migration guide and tooling for moving translations from Crowdin/Phrase/POEditor into Lokalise, including transforms, uploads and validation.