GitHub Repo Briefing
diagram-design
A skill for Claude Code, Codex, and Pi that creates 27 diagram types as static HTML and SVG. Use it for architecture, flows, data models, and timelines; it can also redraw Mermaid or draw.io sources.
A first test
Install the skill, then give it one client process, such as a booking flow. Ask for an architecture or swimlane diagram, open the HTML, and remove any node that does not help the reader.
semantica
A Python library for storing entities, relationships, decisions, rules, and source links in a knowledge graph. It can trace decisions, flag conflicts, and export an audit trail.
A first test
Install it with pip install semantica. Record one vendor choice with its reason and source, then call trace_decision_chain() and check_decision_rules() before using it in a report.
prime-agent
A coding and research agent for long tasks. It keeps a Python session, subagents, goals, schedules, and reusable skills so work can resume after a terminal disconnects.
A first test
Start it inside a disposable clone and give it one bounded task, such as a pull-request review. Check prime-agent status, agents, and attach before accepting changes in the real project.
holehe
A Python CLI that checks whether an email address is linked to accounts on more than 120 services by reading registration or password-recovery responses.
A first test
Install it with pip3 install holehe and test an address that you own or have permission to check. Treat a positive result as a lead, not proof of account ownership, and note rate limits.
Switchyard
A pre-alpha Rust proxy and library for LLM traffic. It translates OpenAI and Anthropic formats, routes requests to configured backends, and records request and latency metrics.
A first test
Install the CLI, create one route, and run switchyard-server with --dry-run. Bind it to 127.0.0.1, call /health, and inspect the backend and metrics for one test request.
code-graph-rag
A CLI that parses a multi-language repository with Tree-sitter and stores its code structure in Memgraph for symbol queries, caller searches, dead-code checks, and edit previews.
A first test
Index a disposable clone with cgr daemon up and cgr start --repo-path /path/to/repo --update-graph. Ask one function question and inspect the source and graph path before allowing an edit.
needle
A 45-million-parameter local model for tool calling and structured extraction. It is distributed in a small file, returns schema-constrained JSON, and supports confidence thresholds.
A first test
Install cactus-needle, define one typed Python tool, and call agent.run() with an invoice example. Accept a device action only above a confidence threshold that you set and test.
macro
An open-source workspace that links email, messages, documents, tasks, calls, agents, and CRM records. Its @links create bidirectional connections between related work.
A first test
Start with one support email. Turn it into a task, link the task to the discussion, and open the original email from the task to verify that the context stays connected.
ToolJet
A visual builder for internal apps with tables, forms, charts, JavaScript, Python, and connections to databases, APIs, SaaS tools, and object storage.
A first test
Run the Docker quickstart with test data and build a small order dashboard. Check the data source, access rules, empty state, and error state before using real records.
TencentDB-Agent-Memory
TencentCloud/TencentDB-Agent-Memory
A self-hosted memory server for agent teams. It turns conversations, documents, and code into Chat Memory, Skills, Wiki, and CodeGraph assets that can be shared by client tools.
A first test
Copy .env.example to .env, run ./start-all.sh, and open localhost:8125. Import one runbook and one codebase, then give a reviewer only the assets it needs.
agent-skills
A pack of 24 workflow skills for definition, planning, implementation, testing, review, performance, simplification, and release work in coding agents.
A first test
Install the pack with npx skills add addyosmani/agent-skills. In one real repo, run /spec, /plan, and /test on a small change, then keep it if the outputs provide useful evidence.
manim
A Python engine for precise programmatic animations. This repository is ManimGL, installed as manimgl; it renders scenes from Python and can use LaTeX for math text.
A first test
Install manimgl with Python 3.10+ and FFmpeg, then run manimgl example_scenes.py OpeningManimExample. Render one short explainer and keep its dependencies separate from Manim Community.