{"slug": "i-built-25-executable-skills-for-ai-coding-agents-all-open-source", "title": "I built 25 executable skills for AI coding agents �“ all open source", "summary": "A developer released 25 open-source executable skills for AI coding agents that fix common problems like token waste, bug multiplication, and context loss. The skills, available for Claude, Cursor, OpenAI Codex, and Hermes Agent, provide step-by-step workflows for systematic debugging, token efficiency, and self-regulation.", "body_md": "**Your AI coding agent keeps doing dumb things? Here's the fix.**\n\n**#1** Agent does`/clear`\n\nand forgets everything — you re-explain from scratch**#2** Agent sees \"fix this bug\" and starts changing code — one fix, three new bugs**#3** Token burns fast — $5 gone before lunch, and you don't know why**#4** Context fills up and agent starts forgetting what you said 5 minutes ago**#5** Agent pushes code changes without asking — you only find out when it breaks\n\n**This pack of 25 executable skills fixes all of that.** Not vague advice. Step-by-step workflows your agent loads and follows.\n\n*Watch the full 30-second quick install — click to play!*\n\n*SVG animation — works everywhere, no loading time, scales perfectly.*\n\n```\ncp -r skills/* ~/.claude/skills/   # Claude Code\ncp -r skills/* ~/.cursor/skills/   # Cursor\ncp -r skills/* ~/.codex/skills/    # OpenAI Codex\n# Hermes Agent: auto-dispatched from ~/.hermes/skills/\n```\n\n**Then tell your agent:** *\"Use the systematic-debugging skill to help me with this error.\"*\n\nAI coding agents are powerful, but they have a fatal flaw: **they don't know how to work safely yet**. Without guidance, an agent will:\n\n- Burn through your token budget in minutes\n- Fix one bug and introduce three more\n- Push code changes without asking\n- Forget context after every\n`/clear`\n\n- Loop forever on the same error\n\nThese 25 skills are executable workflows that teach your agent **how to behave** — systematic debugging, token-aware planning, self-regulation, and safety nets. They're not vague advice; they're step-by-step instructions your agent follows automatically.\n\n| Feature | Hermes Core Skills | LangChain Agents | Manual Prompting |\n|---|---|---|---|\n| Drop-in ready | ✅ Copy and use | ❌ Requires code integration | ❌ Write prompts yourself |\n| Executable workflows | ✅ Agent follows step by step | ❌ Just a framework | ❌ No structure |\n| Cross-platform | ✅ Claude / Codex / Cursor / Hermes | ❌ Python only | ✅ Any agent |\n| Token-aware | ✅ Built-in efficiency rules | ❌ No token optimization | ❌ No token optimization |\n| Self-regulation | ✅ Brake system, stall detection | ❌ Not available | ❌ Not available |\n| Open source | ✅ MIT | ✅ MIT | ✅ Free |\n\nHermes Core Skills is not a framework you integrate — it's a **workflow layer** your agent loads. It works alongside any agent and any framework.\n\n```\nYou: \"Use the systematic-debugging skill on this error: TypeError: Cannot read ...\"\nAgent: [Loads the 4-stage workflow]\n  1. Report — captures the full error context\n  2. Context — reads affected files\n  3. Hypothesis — identifies root cause\n  4. Fix — only applies change after root cause is confirmed\nResult: One fix, zero new bugs.\nYou: \"Follow the token-efficiency skill for this task.\"\nAgent: [Loads token-saving rules]\n  - Compresses long context before proceeding\n  - Delegates heavy research to sub-agents\n  - Avoids wasteful pattern loops\nResult: Uses 40-60% fewer tokens.\nYou: \"Run subagent-driven-development to implement the auth system.\"\nAgent: [Splits into parallel sub-agents]\n  - Sub-agent 1: Login page\n  - Sub-agent 2: JWT middleware\n  - Sub-agent 3: Database schema\n  - Review pass: merges all, fixes conflicts\nResult: Features built in parallel, completed faster.\n```\n\n| Skill | One-liner | Problem it solves |\n|---|---|---|\nsystematic-debugging |\nNo root cause, no fix | Agent randomly patches bugs, making things worse |\nself-regulation-brake-system |\nForce stop after 3 failures | Agent looping forever, burning your budget |\nwriting-plans |\nWrite a plan before touching code | Agent builds the wrong thing, needs redo |\nsubagent-driven-development |\nSplit into subagents, review after | Complex tasks overwhelm a single agent's context |\ntoken-efficiency |\nEvery token counts | End-of-month surprise bills |\ncheckpoints-and-rewind |\nAuto-backup before any change | Agent destroys a file, can't recover |\n\n| Skill | Role | Pain Point | Highlight |\n|---|---|---|---|\nself-regulation-brake-system |\nAgent seatbelt | Agent crashes and keeps burning tokens when you're away | 3-fail stop, 5-min stall report, no bypass allowed |\nsystematic-debugging |\nYour Sherlock Holmes | Agent randomly patches bugs, one fix creates three more | 4-stage: Report → Context → Hypothesis → Fix. Iron rule: no root cause, no fix |\nwriting-plans |\nYour project manager | Agent builds in wrong direction, discovers at the end | Bite-size tasks, exact file paths, code + test per task |\nspec-driven-development |\nYour requirements doctor | Unclear requirements, builds the wrong thing | Write spec first, no coding without understanding |\ntest-driven-development |\nYour quality gate | Agent says \"done\" but never actually tested | RED-GREEN-REFACTOR, no tests = not done |\nsubagent-driven-development |\nYour team lead | Complex task doesn't fit in one agent context | Split tasks → fresh subagent each → review → merge |\nrequesting-code-review |\nYour code reviewer | Agent commits bad code, you don't know | Security scan + quality gate + independent reviewer |\nsecurity-hardening-checklist |\nYour security advisor | Agent doesn't know secure coding, leaves vulnerabilities | Input, auth, storage, third-party, item by item |\nthink-tool |\nYour rational voice | Agent makes impulsive decisions without thinking first | Pros cons + trade-offs + risk analysis framework |\ntoken-efficiency |\nYour CFO | Token burn rate is scary, don't know how to save | Context compression, delegate strategy, waste pattern avoidance |\ncheckpoints-and-rewind |\nYour undo button | Agent corrupts a file, can't roll back | Auto-backup, snapshot, rollback before any change |\ncontext-aware-task-decomposition |\nYour context doctor | Context full, agent starts forgetting | Auto-decompose tasks, never hit context limit |\ncontext-compaction-verification-and-recovery |\nYour memory detective | After compaction, agent doesn't remember what it did | Verify tool commands actually executed |\n\n| Skill | Role | Pain Point | Highlight |\n|---|---|---|---|\nagent-capability-comparison-methodology |\nYour agent buyer | Don't know which agent is good, marketing lies | Source code + benchmark + hands-on, three-layer verification |\nopen-source-adaptation-pattern |\nYour technical due diligence | Install an OSS project, find out it doesn't fit | License + maintenance + community + actual need, four-dimension eval |\nmulti-agent-browser-text-extraction |\nYour research team | JS-heavy sites, browser itself can't extract | Multiple subagents extract in parallel, merge results |\nskill-slimming-strategy |\nYour diet plan | SKILL.md too long, agent loads it and half context is gone | Keep core workflow, move details to references/ |\nbatch-skill-description-standardization |\nYour admin assistant | Dozens of skills with inconsistent descriptions | Fix 100+ at once |\nhermes-improvement-multiphase-plan |\nYour CTO | Want to improve agent but don't know where to start | IDE docs → nightly release → plugin marketplace → desktop app |\n\n| Skill | Role | Pain Point | Highlight |\n|---|---|---|---|\ncross-session-execution-framework |\nYour project continuity | Next session agent doesn't remember what it did | File-based state persistence, recover without memory loss |\nplan |\nYour brake pedal | User says \"plan it first\" but agent starts coding immediately | Pure planning mode, output checklist for approval |\nmulti-role-synthesis-framework |\nYour board of directors | Single-role decisions have blind spots | Multiple roles each give advice → integrated verdict |\n\n| Skill | Role | Pain Point | Highlight |\n|---|---|---|---|\nopenclaw-hermes-arch |\nYour architecture diagram | Don't understand how agent and gateway divide work | Clear responsibility docs + failover mechanism |\nhermes-agent |\nYour Hermes setup guide | New to Hermes, don't know how to set up | Complete zero-to-running guide |\nautonomous-work-signaling |\nYour team coordinator | Multiple autonomous agents don't know what each other is doing | Cross-session work status synchronization |\n\n| Who | Problem | Skill Pack Solution |\n|---|---|---|\nSolo developer |\nAgent burns tokens, context gets lost | `token-efficiency` + `context-aware-task-decomposition` |\nStartup CTO |\nJunior devs using AI produce inconsistent code | `requesting-code-review` + `spec-driven-development` |\nOpen source maintainer |\nNeed help but can't trust AI with security | `security-hardening-checklist` + `systematic-debugging` |\nAgency owner |\nMultiple agents running, no coordination | `autonomous-work-signaling` + `cross-session-execution-framework` |\nAI researcher |\nEvaluating which agent to use for a project | `agent-capability-comparison-methodology` + `open-source-adaptation-pattern` |\n\n| Category | Platform |\n|---|---|\nAI Code Assistants |\nClaude Code, OpenAI Codex CLI, Cursor, Hermes Agent, GitHub Copilot |\nAgent Frameworks |\nLangChain, CrewAI, AutoGen, Any MCP-compatible agent |\nMCP Clients |\nClaude Desktop, VS Code via Continue/Cline, JetBrains, any MCP host |\n\n**New to AI coding agents? Here's everything you need to know.**\n\n-\n**What is an AI coding agent?** A tool like Claude Code or Cursor that can write, edit, and debug code for you in your terminal or editor. -\n**What's a \"skill\"?** A skill is a Markdown file that teaches your agent*how*to do something properly — like a recipe for your AI chef. -\n**How do I use these skills?**- Install with one command (\n`cp -r skills/* ~/.claude/skills/`\n\n) - Tell your agent:\n*\"Use the [skill name] skill\"* - Your agent follows the instructions automatically\n\n- Install with one command (\n-\n**Which skill should I start with?****First**:`systematic-debugging`\n\n— the most useful for daily coding**Second**:`token-efficiency`\n\n— saves you money immediately**Third**:`checkpoints-and-rewind`\n\n— safety net, never lose work\n\nStill confused? [Open a discussion](https://github.com/chrislamlayer1-gif/hermes-core-skills/discussions) — we'll help you get started.\n\nHelp translate this README! Click a badge to contribute:\n\nCurrently English only. Translations welcome — submit a discussion or PR!\n\n- v1.0.0 — 25 core skills released (Jun 2026)\n- v1.1.0 — Skill index (\n`index.json`\n\n) for agent discovery - v1.2.0 — Interactive playground on GitHub Pages\n- v2.0.0 — Community-contributed skills + skill templates\n- Add mappings to common frameworks (pain point categories)\n\n\"Finally — skills that actually tell the agent what to do instead of just giving it vague instructions.\"—Early adopter feedback\n\n\"The self-regulation-brake-system alone saved me from a $50 runaway agent bill.\"—Solo developer, Jun 2026\n\n\"Drop-in ready and zero config. This is what agent tooling should be.\"—Open source contributor\n\n*Have feedback? Open a discussion and share your experience.*\n\n...\n\n*Want to feature this project? Let us know.*\n\nWe'd love your help! Here's how to get started:\n\n**Good first issues (no coding needed):**\n\n- 📝\n**Review a skill**— Try one and open an issue with your feedback - 🌐\n**Translate README**— Pick your language and submit a translation - 🐛\n**Report a bug**— If a skill doesn't work with your agent, let us know - ✨\n**Suggest new skills**— Open a discussion with your pain point\n\n**Code contributions:**\n\n- Browse\n[open issues](https://github.com/chrislamlayer1-gif/hermes-core-skills/issues)— look for`good first issue`\n\nlabels - Fork the repo and create a feature branch\n- Make your changes and submit a PR\n- Your PR will be reviewed within 48 hours\n\nAll contributors are expected to follow our [Code of Conduct](/chrislamlayer1-gif/hermes-core-skills/blob/main/CODE_OF_CONDUCT.md). See the [Contributing Guide](/chrislamlayer1-gif/hermes-core-skills/blob/main/CONTRIBUTING.md) for details.\n\nIf you use this project in your work:\n\n```\n@software{hermes_core_skills,\n  author = {Lam, Chris},\n  title = {Hermes Core Skills},\n  year = {2026},\n  url = {https://github.com/chrislamlayer1-gif/hermes-core-skills},\n  note = {25 executable AI agent skills for debugging, planning, token efficiency, and security. MIT licensed.}\n}\n```\n\nFound a security vulnerability? Please review our [Security Policy](/chrislamlayer1-gif/hermes-core-skills/blob/main/SECURITY.md) for responsible disclosure.\n\nMIT — Use freely, contribute back when you can.", "url": "https://wpnews.pro/news/i-built-25-executable-skills-for-ai-coding-agents-all-open-source", "canonical_source": "https://github.com/chrislamlayer1-gif/hermes-core-skills", "published_at": "2026-06-29 10:19:23+00:00", "updated_at": "2026-06-29 10:29:19.877823+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "ai-products", "developer-tools", "large-language-models"], "entities": ["Claude", "Cursor", "OpenAI Codex", "Hermes Agent", "LangChain", "MIT"], "alternates": {"html": "https://wpnews.pro/news/i-built-25-executable-skills-for-ai-coding-agents-all-open-source", "markdown": "https://wpnews.pro/news/i-built-25-executable-skills-for-ai-coding-agents-all-open-source.md", "text": "https://wpnews.pro/news/i-built-25-executable-skills-for-ai-coding-agents-all-open-source.txt", "jsonld": "https://wpnews.pro/news/i-built-25-executable-skills-for-ai-coding-agents-all-open-source.jsonld"}}