Coding & Development

Portable agent skills: Ideon canvas MCP + MCP Order CLI

Recently updated agent-skill-related GitHub repository: CrownJoker07/AgentSkills.

Privacy-safe Codex skill for battery evidence and mechanism audits | 电池与材料电化学证据、机理及论文结论审查

Recently updated agent-skill-related GitHub repository: PontiusTheBarbarian/claude-skills.

Recently updated agent-skill-related GitHub repository: JhihJian/codex-skills-manager.

Claude skill that forces one deliberate colour palette across generated docs, decks, and spreadsheets, instead of the default rainbow-per-section Tailwind look.

Skills de Claude Code que llevan una idea a produccion vigilada, slice a slice, sin soltar el merge ni el rollback.

Roll a fix to N machines and prove it landed on each one: applied-vs-claimed accounting, canary waves, a verify that must read a fact, and a board that names who is behind. Stdlib-only, MIT.

面向 Codex 的实用开源技能集合

Make Claude Code behave consistently across repos, sessions and humans: rules-as-files governance with precedence, declined-decisions journal, objection sparring. Sanitized from a live production system. Free, MIT.

⚓️ A shared knowledge layer for coding agents. Serves each agent only the decisions and conventions that govern the file it has open, over MCP — then records whether the code still honors them. Your source never leaves your machine.

Graph RAG on SQLite for AI agents: vector retrieval + hand-curated wikilink graph + cross-encoder rerank, with a zero-token per-turn memory ledger. Working pilot.

Improves a production agent system by repairing the evals that judge it. Failures are mined into new eval cases, and gates on both sides control what enters: false negatives the suite would never catch, false positives that refuse a correct change. The system improves by having to pass.

Run AI agents across machines without state drift: consensus protocol (propose/counter/accept/commit), dual-rail message bus, ACK discipline, self-healing sync. Formerly claude-consensus. Free, MIT.

He wrote the codebase twenty years ago. He questions your premise, rejects wrong-file hacks, defends invariants, and fixes the real root cause in 3 lines.

Nobody reviews themselves - and one reviewer model is one blind spot. Fan every change out to several model FAMILIES at once: verdict contract, quorum by family, honest skips, and an exit code that refuses to go green on one vendor. Stdlib-only, MIT.

vibeweaver 的 deepseek harness 专属发行版，帮你的 dsh 交付可信任，经过验证的代码。

Every number you publish is a claim. Prove it in CI: bind each figure in your README to the artifact it came from and fail the build when they drift. No LLM, no API key, no dependencies.

Your LLM reviewer said APPROVE. Did it? A structured verdict contract: prompt rule + parser + exit-code gate in one stdlib file, with the 42 counterexamples that forced every line. Free, MIT.

Your scheduled job says exit 0 — prove it did the work. Three stdlib-only checks: output freshness, silent no-op detection, rollout proof by reading the fact back. Sanitized from a live agent fleet. Free, MIT.

Catch fabricated RAG citations before they reach the user: a zero-token verbatim gate + a burden-of-proof judge, framework-agnostic.

DIN-5008-Briefe aus Markdown, am fertigen PDF nachgemessen. Sollwerte aus Sekundärquellen.

Your agent setup charges rent on every session, before it does any work. Three stdlib-only instruments: what your wiring costs per session, who burned yesterday, and which paid subscriptions are going undrawn. 0 LLM, 0 network, MIT.

Link WhatsApp to Claude (or any MCP client) in ~20 minutes: a live self-refreshing QR page that makes pairing actually work, production patches, and 12 field gotchas. On top of @sjawhar/whatsapp-mcp.