Give Claude a useful map of your work without loading everything at once. Store focused context in small Markdown files, connect them with profile indexes, and let short prompts pull only the files relevant to the task.
Source episode · 25:00
How I AI: Teresa Torres's Claude Code System for Task Management, Automated Research, and 'Lazy' Promptingwith Claire Vo
Teresa opens her LLM context vault, explains why one large CLAUDE.md became noisy, and shows profile indexes that route Claude to small business or personal context files.
A context library of focused files and lightweight indexes that lets Claude find the right background from a short request.
Anthropic AI assistant
Step by step
Follow the sequence once, then adapt the prompts, checks, and handoffs to your own setup.
4 steps
Create an LLM context folder and split durable knowledge into narrow Markdown files. Useful boundaries include one audience, product, writing style, channel, or personal topic per file. Keep each file focused enough that Claude can load it without dragging unrelated information into the session.
Create profile indexes for broad areas such as business and personal context. Each index should describe what is available and link to the specific files Claude may need, without copying their contents.
# Business context index Use this file to locate business context. - Company overview: `business/company-overview.md` - Audience: `business/audience.md` - Writing style: `business/writing-style.md` - Course details: `business/products/course-name.md` - Marketing channels: `business/marketing-channels.md` Load only the files relevant to the current request.
Keep the global CLAUDE.md brief. Tell Claude which profile index to open for business work and which to open for personal questions. Let those indexes route the next choice so an unrelated request never loads the whole library.
For business requests, start with `llm-context/business-profile.md`. For personal requests, start with `llm-context/personal-profile.md`. Follow the selected index and load only the detail files needed for the current task.
Grow the library from real work. At the end of a useful session, ask Claude what durable knowledge should be captured, then decide whether to update an existing file or add one focused file and link it from the right index.
What did we learn in this session that will matter next time? Propose updates to existing context files first. If a new file is needed, keep it focused on one subject and show where it should be linked from the relevant profile index. Do not save transient conversation details.
This iterative process is key. As you use the system, you'll naturally identify gaps in its knowledge, which you can then fill by creating or updating a small context file.
Turn an idea into a PRD, user stories, and a plan.
After the steps
How to recover when the loop fails and where human judgment helps.
Recover
Keep building
Create context-aware guardrails for your enterprise AI agent using a 'Project' governance layer. As a project owner, define rules, scope tools, and require human-in-the-loop approval for sensitive actions, ensuring safe and effective AI use.
Transform a successful AI chat session into a reusable skill that anyone in your company can use. Package prompts, tools, and code into a standardized workflow, making it easy to share and scale automation across your organization.
Use a natural language AI agent to instantly generate and iterate on custom data dashboards. This workflow queries your company's data securely, allowing non-technical users to perform their own 'last mile' data analysis without filing tickets.
Join 100,000+ product managers who use ChatPRD to write better docs, align teams faster, and build products users love.