Telexed

AI news through a solo-operator lens — only what changes your day25 of 843

FILTER[All][Agents & tools][Models & API][Generative media][Infra & SaaS][ASO & growth][Indie business][Idea signals][Other][★6+ high-signal]

GitHub Trending Weekly ✕clear filters

Yesterday1 dispatches

#0843
#0843Agents & tools GitHub Trending Weekly14 hours ago
`ECC`, cross-harness operator system for coding agents
70radar
ECCAgent ops framework — shared workflows across coding harnesses
Reusable skills, hooks, memory, evals, and security rules target Codex, Claude Code, Cursor, and more. Useful as an agent workflow template, less as a plug-and-play product.
- Claims 182K+ stars, 28K+ forks, and 170+ contributors; adoption signal is strong enough to inspect, even if the number needs separate due diligence.
- Covers skills, instincts, memory optimization, continuous learning, verification loops, subagents, MCP configs, and legacy command shims across agent harnesses.
- ECC Pro adds a hosted GitHub App for private repos at $19/seat/mo; the OSS layer remains MIT-licensed.
- The practical value is pattern extraction: prompts, hooks, eval loops, and repo rules can be copied into existing Codex or Claude Code setups.
Source: github.com/affaan-m/ECCRead original →
FIG-8431:1
70radar
FIG-8431:1

Fri, May 295 dispatches

#0842
#0842Agents & tools GitHub Trending Weekly2 days ago
`CCX`, unified proxy for Claude, Codex, OpenAI Images, and Gemini APIs
70radar
CCXAI API proxy — multi-provider routing and failover
One gateway normalizes multiple AI APIs behind one endpoint. Built-in routing, failover, key rotation, and admin UI make multi-provider agent stacks easier to run.
- Single-port backend serves the web UI, admin API, Claude Messages, OpenAI Chat, Codex Responses, OpenAI Images, and Gemini routes.
- Scheduling includes priorities, promotion windows, health checks, failover, and circuit recovery; useful when one provider stalls.
- Per-channel key rotation, model allowlists, custom headers, proxy support, and route prefixes reduce glue code around Claude Code or Codex setups.
- Responses session tracking covers multi-turn workflows, so agent runs can keep state while switching through a unified gateway.
Source: github.com/BenedictKing/ccxRead original →
FIG-8421:1
70radar
FIG-8421:1
#0841
#0841Agents & tools GitHub Trending Weekly2 days ago
Microsoft `Agent Governance Toolkit` for production agent controls
80radar
Agent Governance ToolkitAI agent governance toolkit — pre-execution policy enforcement
Deterministic policy checks sit between model intent and real tool calls. Useful when agents touch email, DBs, or multi-agent workflows.
- AGT intercepts every tool call, message send, and delegation in app code before execution; denied actions become structurally impossible.
- Covers 10/10 OWASP Agentic Top 10 with policy enforcement, zero-trust identity, sandboxing, and SRE-oriented reliability controls.
- The audit trail records active policy, requested action, and allow/deny reason, which matters once agents operate against real customer or business data.
- Install path is simple: pip install agent-governance-toolkit[full]; still Public Preview, so pin versions before wiring it into production.
Source: github.com/microsoft/agent-governance-toolkitRead original →
FIG-8411:1
80radar
FIG-8411:1
#0840
#0840Agents & tools GitHub Trending Weekly2 days ago
`Taste-Skill`, portable design skills for AI coding agents
60radar
Taste-SkillAgent skill pack — improves design quality of AI-built UIs
Adds reusable frontend taste rules to agents like Codex, Cursor, and Claude Code. Useful for reducing generic AI UI output before handoff.
- Install all skills with npx skills add https://github.com/Leonxlnx/taste-skill; single-skill install uses --skill "design-taste-frontend".
- Default design-taste-frontend is now v2 experimental, with brief inference, design-system mapping, GSAP skeletons, and redesign-audit checks.
- Includes image-generation skills for web, mobile, and brand-kit reference boards; they output frames, then coding agents implement them.
- design-taste-frontend-v1 remains available for pinning when exact old behavior matters.
Source: github.com/Leonxlnx/taste-skillRead original →
FIG-8401:1
60radar
FIG-8401:1
#0839
#0839Agents & tools GitHub Trending Weekly2 days ago
Anthropic opens official `Claude Code Plugins` directory
80radar
Claude Code PluginsCoding agent plugin system — packages MCP, skills, and commands
Plugin discovery now has an Anthropic-managed marketplace. It makes reusable agents, skills, slash commands, and MCP integrations easier to package, but install trust still matters.
- Install flow is built into Claude Code: run /plugin install {plugin-name}@claude-plugins-official or use /plugin > Discover.
- Plugin packages can include .mcp.json, commands/, agents/, and skills/; this turns workflow pieces into reusable tooling.
- External plugins must pass quality and security review, yet Anthropic says it does not control plugin contents after install.
- Skill bundles can expose selected SKILL.md directories without a strict manifest, useful for shipping curated agent capabilities from existing repos.
Source: github.com/anthropics/claude-plugins-officialRead original →
FIG-8391:1
80radar
FIG-8391:1
#0838
#0838Agents & tools GitHub Trending Weekly2 days ago
`Anthropic Cybersecurity Skills`: 754 security skills for AI agents
60radar
Anthropic Cybersecurity SkillsAgent security skill library — maps to five frameworks
A reusable agentskills.io library gives coding agents structured security playbooks. Useful for audits and incident triage, but trust it as guidance, not authority.
- Covers 754 skills across 26 security domains, mapped to ATT&CK, NIST CSF, ATLAS, D3FEND, and AI RMF.
- Works with Claude Code, Codex CLI, Cursor, Gemini CLI, and other agentskills.io platforms, so adoption cost is low.
- Quick start is npx skills add mukul975/Anthropic-Cybersecurity-Skills; useful as a reusable checklist layer for security tasks.
- It is an independent community project, not Anthropic-official. Review skill quality before using it in production workflows.
Source: github.com/mukul975/Anthropic-Cybersecurity-SkillsRead original →
FIG-8381:1
60radar
FIG-8381:1

Wed, May 278 dispatches

#0837
#0837Agents & tools GitHub Trending Weekly3 days ago
`dotnet/skills`: Official .NET Agent Skills Marketplace
60radar
dotnet/skillsAgent skills repo — curated by the .NET team
Microsoft’s .NET team is packaging task-specific skills for coding agents. Useful if your stack touches C#, but narrow outside .NET work.
- Covers dotnet, dotnet-data, dotnet-diag, dotnet-msbuild, dotnet-nuget, dotnet-test, dotnet-aspnet, dotnet-blazor, and more.
- Installation targets Copilot CLI, Claude Code, VS Code, and Cursor, so the distribution model is agent plugin marketplace, not just docs.
- Skill areas include build failures, EF data access, NuGet modernization, diagnostics, testing, upgrades, ASP.NET Core, Blazor, MAUI, and LLM integration.
- The dashboard tracks accuracy and efficiency scoring trends, making this closer to a maintained agent toolchain than a prompt dump.
Source: github.com/dotnet/skillsRead original →
FIG-8371:1
60radar
FIG-8371:1
#0836
#0836Agents & tools GitHub Trending Weekly3 days ago
`claude-code-harness`, a Plan→Work→Review Loop for `Claude Code`
70radar
claude-code-harnessCoding agent harness — file-based plan, work, review loop
Turns agent coding into a gated delivery flow: spec first, approved slice next, verification before release. Useful when raw chat-based coding keeps drifting.
- /harness-plan generates spec.md and Plans.md with scope, acceptance criteria, unknowns, and stop conditions; planning moves out of chat.
- /harness-work limits implementation to the approved slice and expects TDD plus verification, reducing uncontrolled agent edits.
- /harness-review separates review from implementation; this is a practical guardrail for solo PR quality.
- Install path is short: marketplace add, plugin install, then /harness-setup; the first usable test can be one small README task.
Source: github.com/Chachamaru127/claude-code-harnessRead original →
FIG-8361:1
70radar
FIG-8361:1
#0835
#0835Agents & tools GitHub Trending Weekly3 days ago
Anthropic Open-Sources `Knowledge Work Plugins` for `Claude Cowork`
60radar
Knowledge Work PluginsClaude plugin repo — role-based agent workflow bundles
Role-specific plugin bundles combine skills, connectors, slash commands, and sub-agents. The repo is more useful as a template for repeatable AI workflows than as a plug-and-play solo stack.
- The repo ships 11 plugins across productivity, sales, support, PM, marketing, legal, finance, data, and enterprise search.
- Each plugin bundles skills, connectors, slash commands, and sub-agents, giving a concrete structure for packaging repeatable agent workflows.
- Connector lists skew enterprise: Slack, Notion, Linear, Jira, Figma, HubSpot, BigQuery, and Microsoft 365 dominate.
- Best takeaway is the plugin architecture: encode role context, tool access, and workflow commands instead of rewriting prompts every time.
Source: github.com/anthropics/knowledge-work-pluginsRead original →
FIG-8351:1
60radar
FIG-8351:1
#0834
#0834Agents & tools GitHub Trending Weekly4 days ago
`chrome-devtools-mcp` brings Chrome DevTools to coding agents
80radar
chrome-devtools-mcpMCP server — exposes Chrome DevTools to coding agents
Coding agents can now inspect live Chrome through MCP: traces, network, screenshots, console, and Puppeteer actions. Useful now for frontend debugging, but isolate sensitive sessions.
- Runs as an MCP server for Antigravity, Claude, Cursor, Copilot, and similar agents; also ships a CLI for non-MCP use.
- Performance tooling records Chrome traces and can enrich lab results with CrUX field data; disable that path with --no-performance-crux.
- Debug surface includes network requests, screenshots, and source-mapped console stack traces, so agents can diagnose UI failures closer to a real browser.
- Sensitive browser contents are exposed to connected MCP clients. Use a separate Chrome profile, test account, or throwaway session.
- Usage statistics are on by default; opt out with --no-usage-statistics or CHROME_DEVTOOLS_MCP_NO_USAGE_STATISTICS.
Source: github.com/ChromeDevTools/chrome-devtools-mcpRead original →
FIG-8341:1
80radar
FIG-8341:1
#0833
#0833Agents & tools GitHub Trending Weekly4 days ago
`Routa`: Workspace-First Multi-Agent Board for Software Delivery
70radar
RoutaMulti-agent dev platform — board-based task and review orchestration
Agent work moves out of one chat thread into a shared board. Useful for longer coding jobs where tasks, evidence, traces, and review gates must stay visible.
- Routa uses a Kanban board as the coordination bus, with separate specialist prompts and evidence contracts per lane.
- Web and desktop share one contract via api-contract.yaml; Next.js, Tauri, Axum, REST, and SSE sit behind the same semantics.
- Review is split into trace monitoring, policy/evidence checks, and gate verification, reducing blind trust in one reviewer persona.
- Supports MCP, ACP, A2A, AG-UI, and A2UI, so it targets multi-agent orchestration rather than another single coding chat.
Source: github.com/phodal/routaRead original →
FIG-8331:1
70radar
FIG-8331:1
#0832
#0832Agents & tools GitHub Trending Weekly4 days ago
`Cursor` Official Plugin Spec and Marketplace Repo
80radar
CursorAI coding IDE — agent and plugin workflows
Plugins now have a concrete manifest, skills, rules, and MCP layout. The useful part is not the list itself, but the repeatable packaging pattern for agent workflows.
- Each plugin is a standalone root directory with .cursor-plugin/plugin.json, making distribution and validation manifest-driven.
- The catalog covers memory updates, PR review canvases, docs canvases, SDK integrations, and parallel agent orchestration.
- create-plugin scaffolds and validates new plugins, lowering the setup cost for reusable internal workflows.
- orchestrate points to a planner-worker-verifier pattern for large tasks, useful when one agent session gets too wide.
Source: github.com/cursor/pluginsRead original →
FIG-8321:1
80radar
FIG-8321:1
#0831
#0831Agents & tools GitHub Trending Weekly4 days ago
`CLI-Anything` turns desktop software into agent-ready CLIs
80radar
CLI-AnythingAgent CLI hub — wraps GUI software for command-line control
Agent workflows can now drive GUI-heavy tools through generated command-line harnesses. The useful angle is broader than coding: media, CAD, documents, and app automation become scriptable.
- CLI-Hub installs community CLIs with pip install cli-anything-hub and cli-hub install <name>; discovery and setup are packaged as one flow.
- Recent merged harnesses cover Rekordbox, Calibre, 3MF, MiniMax, and UEAtelier, pushing agents into media, ebooks, 3D, TTS, and Unreal workflows.
- Security hardening matters: untrusted XML/SVG/ODF/MLT/MusicXML/CSL parsing now routes through defusedxml, reducing plugin-style automation risk.
- The strongest signal is the SKILL.md unification under top-level skills/; CLIs are being packaged as reusable agent capabilities, not one-off scripts.
Source: github.com/HKUDS/CLI-AnythingRead original →
FIG-8311:1
80radar
FIG-8311:1
#0830
#0830Agents & tools GitHub Trending Weekly4 days ago
`Orca`, an IDE for running multiple coding agents in parallel
80radar
OrcaCoding agent IDE — parallel CLI agent runs
Bring your own subscriptions and run CLI agents side by side across worktrees. Strong fit for parallel repo work if agent cost is already budgeted.
- Supports Claude Code, Codex, Grok, Gemini, OpenCode, and other CLI agents, so it avoids locking the workflow to one vendor.
- Each task gets its own git worktree; fewer stash and branch conflicts when testing multiple agent outputs in parallel.
- Built-in diffs, quick edits, commits, GitHub links, and Actions checks turn agent runs into reviewable work, not loose terminal sessions.
- SSH support and mobile companion apps expand usage beyond a single laptop. Useful when long-running agents need remote monitoring.
Source: github.com/stablyai/orcaRead original →
FIG-8301:1
80radar
FIG-8301:1

Tue, May 263 dispatches

#0829
#0829Agents & tools GitHub Trending Weekly5 days ago
`oh-my-pi`: terminal coding agent with LSP, debugger, and subagents
70radar
oh-my-piTerminal coding agent — built-in LSP and debugger
A CLI agent bundles IDE-grade context, execution, browser, and debugger tools out of the box. Worth testing if current agent edits fail often.
- Ships with 40+ providers, 32 built-in tools, 13 LSP ops, and 27 DAP ops; less glue code around a coding agent stack.
- Hash-anchored edits claim a Grok Code Fast 1 pass-rate jump from 6.7% to 68.3%; edit format quality is the pitch.
- Persistent Python and Bun workers can call agent tools such as read, search, and task; useful for data-heavy debugging loops.
- LSP rename and debugger attach paths cover imports, barrels, lldb, dlv, and debugpy; closer to IDE automation than a plain shell bot.
Source: github.com/can1357/oh-my-piRead original →
FIG-8291:1
70radar
FIG-8291:1
#0828
#0828Agents & tools GitHub Trending Weekly5 days ago
`Understand Anything`: interactive codebase knowledge graph for coding agents
70radar
Understand AnythingCode understanding tool — turns repos into interactive knowledge graphs
Turns large repos into searchable architecture maps with tours and impact analysis. Useful when onboarding or refactoring unfamiliar code, especially before handing work to agents.
- Builds a graph of files, functions, classes, and dependencies, then exposes summaries and relationships in a dashboard.
- Works alongside Claude Code, Codex, Cursor, Copilot, and Gemini CLI; positioning is agent companion, not another IDE.
- Diff Impact Analysis shows affected system areas before commit. That is the most practical hook for risky refactors.
- Domain view maps code to business flows, making legacy SaaS logic easier to inspect before changing billing, auth, or data paths.
Source: github.com/Lum1104/Understand-AnythingRead original →
FIG-8281:1
70radar
FIG-8281:1
#0827
#0827Agents & tools GitHub Trending Weekly5 days ago
`Scientific Agent Skills`: 138 Open Skills for Research Agents
60radar
Scientific Agent SkillsAgent skill collection — 138 scientific workflows
Research workflows are packaged as reusable agent Skills, not one-off prompts. Strong reference material for vertical agent products, though most value sits in science-heavy niches.
- 138 skills cover genomics, drug discovery, medical imaging, time series, physics, finance, and writing; breadth matters more than any single workflow.
- Works with Cursor, Claude Code, Codex, and other agents using the open Agent Skills standard; portability is the useful signal.
- K-Dense BYOK adds a local desktop research workspace with BYO keys, 40+ models, web search, file handling, and 100+ scientific databases.
- The bigger takeaway is packaging domain workflows as installable skills. Good pattern for narrow SaaS agents with docs, examples, and tool recipes.
Source: github.com/K-Dense-AI/scientific-agent-skillsRead original →
FIG-8271:1
60radar
FIG-8271:1

Fri, May 221 dispatches

#0826
#0826Agents & tools GitHub Trending Weeklylast week
`12-factor-agents`: production patterns for reliable LLM apps
60radar
A practical checklist for turning LLM steps into maintainable software. Strong fit when an agent feature needs shipping discipline, not another framework.
- Core stance: good agents are mostly deterministic software with LLM calls placed where they add leverage.
- Own prompts, context windows, control flow, and execution state. Treat frameworks as optional, not the architecture.
- Tools become structured outputs, and errors get compacted back into context. This maps well to production debugging.
- Small focused agents plus pause/resume APIs reduce blast radius. Useful for SaaS workflows with human approval steps.
Source: github.com/humanlayer/12-factor-agentsRead original →
FIG-8261:1
60radar
FIG-8261:1

Thu, May 211 dispatches

#0825
#0825Agents & tools GitHub Trending Weeklylast week
`anthropics/skills` Opens Claude Agent Skill Examples and Spec
70radar
Agent SkillsAgent task package standard — bundles instructions, scripts, resources
Reusable task packs are becoming a practical agent interface. The repo gives concrete SKILL.md, script, and resource patterns worth copying.
- Each skill is a folder with SKILL.md, scripts, and resources, loaded dynamically by Claude for repeatable specialized work.
- The repo includes examples across creative, technical, enterprise, and document workflows, plus ./spec and ./template for custom skills.
- Document skills for docx, pdf, pptx, and xlsx are source-available references, not Apache-licensed open source.
- Claude Code can register it as a plugin marketplace with /plugin marketplace add anthropics/skills, lowering setup friction.
Source: github.com/anthropics/skillsRead original →
FIG-8251:1
70radar
FIG-8251:1

Wed, May 202 dispatches

#0824
#0824Agents & tools GitHub Trending Weekly2 weeks ago
`OpenHuman`: open-source desktop AI agent with memory and 118+ integrations
70radar
OpenHumanDesktop AI agent — local memory with 118+ integrations
A UI-first personal agent bundles OAuth integrations, local memory, coder tools, voice, and model routing. Early beta, but worth watching if it turns app context into reusable local knowledge.
- 118+ integrations connect Gmail, Notion, GitHub, Slack, Stripe, Calendar, Drive, Linear, and Jira through one-click OAuth.
- The memory layer canonicalizes data into Markdown chunks, stores summary trees in local SQLite, and mirrors them to an Obsidian-compatible vault.
- Built-in tools cover web search, scraping, filesystem, git, lint, test, grep, STT, ElevenLabs TTS, and Google Meet participation.
- TokenJuice compresses tool output, HTML, URLs, emails, and scrape payloads before LLM calls, reducing context waste.
Source: github.com/tinyhumansai/openhumanRead original →
FIG-8241:1
70radar
FIG-8241:1
#0823
#0823Agents & tools GitHub Trending Weekly2 weeks ago
`CodeGraph`: local code knowledge graph for coding agents
80radar
CodeGraphCode graph CLI — local symbol and call-graph indexing
Pre-indexing gives agents a queryable map of symbols, calls, and structure. The benchmark claims 92% fewer tool calls and faster exploration, so it is worth testing on large repos.
- Runs via npx @colbymchenry/codegraph, then codegraph init -i per project; setup targets Claude Code, Cursor, Codex CLI, and opencode.
- Benchmarks across 6 repos report 92% fewer tool calls and 71% faster exploration on average; strongest gains appear in TypeScript and Swift repos.
- The graph is 100% local, which matters for private codebases where sending extra context or relying on hosted indexing is a non-starter.
- Claimed workflow replaces repeated grep, glob, and file reads with codegraph_explore; less exploration overhead means more budget left for actual edits.
Source: github.com/colbymchenry/codegraphRead original →
FIG-8231:1
80radar
FIG-8231:1

Mon, May 184 dispatches

#0822
#0822Agents & tools GitHub Trending Weekly2 weeks ago
`react-doctor` scores React code quality from the CLI
80radar
react-doctorReact diagnostics CLI — outputs a 0-100 quality score
A single npx run returns a 0-100 score plus concrete diagnostics. Useful as a cheap CI guardrail for agent-written React.
- npx -y react-doctor@latest . scans Next.js, Vite, and React Native projects with framework-aware rules.
- Checks cover state/effects, performance, architecture, security, accessibility, and dead code. That maps well to common agent regressions.
- The included GitHub Action can comment on PRs and expose a score output. Set fail-on to block low-quality changes.
- react-doctor.config.json supports rule, file, and override ignores. Use narrow overrides to keep coverage intact.
- install writes React best-practice guidance into Claude Code, Cursor, Codex, OpenCode, and 50+ agents.
Source: github.com/millionco/react-doctorRead original →
FIG-8221:1
80radar
FIG-8221:1
#0821
#0821Agents & tools GitHub Trending Weekly2 weeks ago
`Bun`: one JavaScript toolchain for runtime, bundling, tests, and packages
80radar
BunJavaScript toolchain — runtime and package manager in one
One executable covers node, package install, scripts, tests, and TS/JSX execution. Fewer moving parts make small app stacks faster to set up and easier to keep lean.
- bun run index.tsx supports TypeScript and JSX out of the box, cutting the usual transpiler setup for quick prototypes.
- The CLI includes bun test, bun install, bunx, and script running, so one binary can replace several Node.js dev dependencies.
- Built in Zig on JavaScriptCore, it targets lower startup time and memory use than typical Node-based tooling.
- Linux, macOS, Windows, Docker, npm, and Homebrew installs are covered; adoption friction is low for existing projects.
Source: github.com/oven-sh/bunRead original →
FIG-8211:1
80radar
FIG-8211:1
#0820
#0820Agents & tools GitHub Trending Weekly2 weeks ago
`Superpowers`: skill framework for coding agents
60radar
SuperpowersCoding-agent skills framework — enforces spec-to-TDD workflow
A reusable skills layer turns coding agents into spec-plan-TDD executors. Useful if your agent often jumps into code too early.
- Runs on Claude Code, Codex CLI, Codex App, Cursor, Copilot CLI, and others — not tied to one agent harness.
- Workflow is explicit: clarify intent, produce readable spec chunks, get design sign-off, then generate an implementation plan.
- The method pushes red/green TDD, YAGNI, and DRY; its value is guardrails, not a stronger base model.
- Subagent-driven execution can run for hours when the plan is tight. Best fit: scoped refactors, feature slices, test expansion.
Source: github.com/obra/superpowersRead original →
FIG-8201:1
60radar
FIG-8201:1
#0819
#0819Agents & tools GitHub Trending Weekly2 weeks ago
`Pyrefly`: Meta’s Fast Python Type Checker and Language Server
80radar
PyreflyPython type checker — fast checks plus IDE features
Python projects get fast type checks plus IDE features in one tool. Strong fit for typed backends and AI-heavy codebases where slow feedback kills iteration.
- Claims 1.85M+ lines/sec and checks PyTorch 15x faster than Mypy and Pyright; save-time IDE rechecks usually finish under 10ms.
- Runs as both CLI and language server, covering navigation, autocomplete, hover info, inlay hints, and semantic highlighting with consistent results.
- Built-in Pydantic and Django support reduces false friction for typical SaaS backends and API projects.
- Migration path is practical: pyrefly init, pyrefly suppress, and pyrefly infer let teams start with one file instead of a full rewrite.
Source: github.com/facebook/pyreflyRead original →
FIG-8191:1
80radar
FIG-8191:1

`ECC`, cross-harness operator system for coding agents

`CCX`, unified proxy for Claude, Codex, OpenAI Images, and Gemini APIs

Microsoft `Agent Governance Toolkit` for production agent controls

`Taste-Skill`, portable design skills for AI coding agents

Anthropic opens official `Claude Code Plugins` directory

`Anthropic Cybersecurity Skills`: 754 security skills for AI agents

`dotnet/skills`: Official .NET Agent Skills Marketplace

`claude-code-harness`, a Plan→Work→Review Loop for `Claude Code`

Anthropic Open-Sources `Knowledge Work Plugins` for `Claude Cowork`

`chrome-devtools-mcp` brings Chrome DevTools to coding agents

`Routa`: Workspace-First Multi-Agent Board for Software Delivery

`Cursor` Official Plugin Spec and Marketplace Repo

`CLI-Anything` turns desktop software into agent-ready CLIs

`Orca`, an IDE for running multiple coding agents in parallel

`oh-my-pi`: terminal coding agent with LSP, debugger, and subagents

`Understand Anything`: interactive codebase knowledge graph for coding agents

`Scientific Agent Skills`: 138 Open Skills for Research Agents

`12-factor-agents`: production patterns for reliable LLM apps

`anthropics/skills` Opens Claude Agent Skill Examples and Spec

`OpenHuman`: open-source desktop AI agent with memory and 118+ integrations

`CodeGraph`: local code knowledge graph for coding agents

`react-doctor` scores React code quality from the CLI

`Bun`: one JavaScript toolchain for runtime, bundling, tests, and packages

`Superpowers`: skill framework for coding agents

`Pyrefly`: Meta’s Fast Python Type Checker and Language Server