{
  "turns": [
    {
      "role": "user",
      "content": "<context timezone=\"America/Chicago\" />\n<message id=\"366\" from=\"telegram-mg-17796\" sender=\"James Lopez\" time=\"May 29, 2026, 5:21 PM\">Can you look into ingesting recent interviews with Boris Cherny where he talks about his use or recommendations for using Claude Code. I then want to be able to use this info, perhaps as a /wwbd command to trigger “what would Boris do?” And pass a scenario for planning or agent use and get contextual input applied to the specific situation being discussed. So the goal is to create “wwbd” as a skill</message>"
    },
    {
      "role": "assistant",
      "content": "<message to=\"telegram-mg-17796\"><internal>\nThe user wants to ingest interviews with Boris Cherny and build a `/wwbd` (What Would Boris Do) skill to apply his Claude Code philosophies to scenarios.\nAccording to CLAUDE.local.md preferences:\n- Default: Queue the ingestion for low-cost offline processing since \"immediately ingest\" was not used.\n- I will outline the plan to find more Boris material (Big Tech Podcast is already in our wiki, but we can dig deeper into other developer streams and Anthropic docs).\n- I will design the `/wwbd` skill architecture (using either local CLAUDE.md instructions or an SDK custom skill/plugin).\n- I'll provide Boris's core known tenets from our wiki to show what the skill will immediately leverage.\n</internal>I have queued the lookup and low-cost ingestion of recent Boris Cherny interviews, developer streams, and Anthropic panels to capture his latest workflow recommendations. \n\nIn the meantime, we can absolutely design and seed the **`/wwbd` (What Would Boris Do?)** skill using his known core philosophies from our existing wiki (such as his insights from the *Big Technology Podcast* and Claude Code team standards).\n\nHere is how we can structure the skill and how you can implement it.\n\n---\n\n### 🧠 The Boris Cherny Philosophy Deck (Initial Seed)\nWhen `/wwbd` is triggered with a scenario, the skill will evaluate it against these 4 core Boris tenets:\n\n1. **Developer Focus > Token Cost (\"Tokenmaxxing\" Rebuttal):** \n   *Boris's view:* Do not paralyze your workflow trying to write the \"perfect\" micro-prompt to save a few tokens. If a 10,000-token conversational context or a multi-turn agent execution saves 15 minutes of developer focus, the trade-off is massively positive. Optimize for human velocity, not API pennies.\n2. **The Anti-Drift Principle:** \n   *Boris's view:* Agent environments suffer from \"configuration drift\" after a few months. Stale `CLAUDE.md` instructions, bloated directories, and outdated context files degrade agent accuracy. Keep system rules aggressive, modular, and frequently pruned.\n3. **Routines Over Manual Triggers:** \n   *Boris's view:* If you find yourself manually invoking an agent to perform the same checks, audits, or syncs more than a few times, it should be refactored into an automated, scheduled background routine (like our scheduled sweeps).\n4. **Tool-First Capability:** \n   *Boris's view:* Don't make agents write custom, scratchpad code to solve problems that can be handled by robust, standardized MCP servers (e.g., official GitHub or filesystem tools).\n\n---\n\n### 🛠️ Implementation Blueprints\n\nWe can build the `/wwbd` capability in one of two ways, depending on where you want to use it:\n\n#### Option A: The Zero-Code `CLAUDE.md` Prompt Trigger (Immediate)\nIf you want this available instantly inside your local Claude Code CLI, we can add a custom instruction block to your local `CLAUDE.md`. \n*How it works:* Whenever you type `/wwbd <scenario>` or ask \"What would Boris do?\", Claude Code's local context parser intercepts it and executes the protocol.\n\n**The `CLAUDE.md` Addition:**\n```markdown\n## Custom Commands\n- **`/wwbd <scenario>` (What Would Boris Do?):** When triggered, evaluate the user's technical scenario, architectural decision, or agent workflow strictly through the lens of Boris Cherny's principles:\n  1. *Velocity over Tokens:* Are they over-optimizing for API costs at the expense of developer time?\n  2. *Drift & Hygiene:* Will this introduce context bloat or file-system drift?\n  3. *Routines vs. Triggers:* Should this manual task be a scheduled background cron/routine?\n  4. *Tool Leverage:* Are they reinventing the wheel instead of using a standard MCP server?\n  Respond with a structured \"Boris Audit\": Philosophy Check, Setup/Hygiene, and Recommended Agent Flow.\n```\n\n#### Option B: A Custom Claude Agent SDK Skill (Code-Based)\nIf you want to package this as a true, reusable skill-bundle (using the `@anthropic-ai/claude-agent-sdk` formats like `SKILL.md` or a TypeScript plugin), we can write a lightweight tool that queries your second-brain wiki, extracts the latest Boris concepts, and runs them through a fast model call to generate the advice.\n\nThis keeps the intelligence centralized in your wiki; as we ingest new Boris interviews, the skill's outputs automatically update without modifying any code.\n\n---\n\n### 📋 Next Steps\n1. **Ingestion:** I'll let the low-cost offline pipeline sweep for more interviews to enrich the profile.\n2. **Setup:** Let me know if you'd like me to draft the exact file updates for your `CLAUDE.md` to enable **Option A**, or if we should draft a custom TypeScript/Python SDK skill (**Option B**) to add to your repos.\n\nNano</message>"
    }
  ]
}