GenAI PMClaude.md
A steering file used to guide Claude Code behavior through repository-specific instructions. It is part of a broader control surface for agent workflows.
Key Highlights
- Claude.md is a repository-specific steering file that makes Claude Code behavior more consistent across projects.
- It has emerged as part of a larger control surface that includes skills, hooks, rules, and subagents.
- Practitioners report that lean, high-signal Claude.md files work better than over-engineered instruction sets.
- Teams are beginning to treat Claude.md as a living operational asset that can be updated from workflow learnings.
- For AI PMs, Claude.md is a practical tool for standardizing agent behavior, quality expectations, and governance.
Claude.md
Overview
Claude.md is a repository-level steering file used to shape how Claude Code behaves inside a specific project. Rather than relying only on prompts in a single session, teams can encode persistent instructions, conventions, constraints, and workflow guidance directly into the repo so the coding agent operates with more consistent context. In practice, Claude.md acts like an operational handbook for the agent: it can tell Claude how to structure changes, what rules to follow, what tools or patterns to prefer, and when to apply project-specific guidance.For AI Product Managers, Claude.md matters because it turns agent behavior from something ad hoc into something designable. It is part of a broader control surface that now includes skills, hooks, rules, and subagents, making it possible to operationalize product, engineering, and quality expectations in reusable form. The emerging lesson from practitioners is that the most effective Claude.md files are not bloated knowledge dumps, but lean, high-signal instruction sets that evolve alongside the model and the team’s workflows.
Key Developments
- 2026-01-13: Bolt added support for Claude.md imports, enabling agents to pull in project context quickly without manual setup.
- 2026-03-18: HumanLayer published guidance on getting Claude to reliably read Claude.md, including the use of conditional XML blocks to surface the right instructions at the right time.
- 2026-06-01: Josh Pigford described a workflow where a `/learnings` skill converts transcripts and code changes into updates for Claude.md, treating the file as a living source of operational guidance.
- 2026-06-18: Jason Zhou argued that Claude.md assets depreciate as models improve, recommending teams periodically clear and rewrite them to preserve only the instructions that materially change behavior.
- 2026-06-19: Anthropic formally highlighted Claude.md as part of new steering controls for Claude Code alongside skills, hooks, rules, and subagents.
Relevance to AI PMs
- Standardize agent behavior across projects: AI PMs can use Claude.md to encode product constraints, codebase norms, testing expectations, and review criteria so agent outputs align more closely with team standards.
- Create repeatable AI workflows: When paired with skills, hooks, and rules, Claude.md helps PMs turn one-off prompting tactics into reusable operating procedures for building, debugging, documentation, and QA.
- Improve governance without slowing teams down: Claude.md offers a lightweight way to express what the agent should and should not do in a repository, helping teams reduce drift, overreach, and inconsistent execution.
Related
- Claude / Claude Code: Claude.md is specifically tied to guiding Claude Code behavior within a repository.
- Skills, hooks, rules, subagents: These are adjacent steering mechanisms that extend Claude.md into a broader agent control framework.
- HumanLayer: Published practical patterns for making Claude.md instructions more reliably attended to.
- Bolt: Added support for importing Claude.md, making repository context more portable across tools.
- Conductor: Appears in adjacent autonomous AI workflows where Claude.md can serve as persistent steering context.
- Josh Pigford: Shared examples of updating Claude.md through a `/learnings` workflow.
- Jason Zhou: Emphasized pruning and rewriting Claude.md as models evolve.
- tal-raviv: Related entity in the broader ecosystem around agent workflows and steering patterns.
Newsletter Mentions (5)
“Announces new steering controls for Claude Code, including CLAUDE.md files and mechanisms like skills, hooks, rules, and subagents to guide agent behavior.”
📝 Claude Code Blog Steering Claude Code: CLAUDE.md files, skills, hooks, rules, subagents and more - Announces new steering controls for Claude Code, including CLAUDE.md files and mechanisms like skills, hooks, rules, and subagents to guide agent behavior.
“Jason Zhou warns that CLAUDE.md assets depreciate as the model evolves, so he routinely clears and rewrites them—keeping only the few rules that truly steer behavior—and argues that over-engineering, not lack of context, is most people’s real issue.”
#21 𝕏 Jason Zhou warns that CLAUDE.md assets depreciate as the model evolves, so he routinely clears and rewrites them—keeping only the few rules that truly steer behavior—and argues that over-engineering, not lack of context, is most people’s real issue.
“He uses a “/but-for-real” skill to coerce the AI into rechecking its own output for additional errors and a “/learnings” skill that distills session transcripts and code changes into updates for his CLAUDE.md guidelines.”
#3 ▶️ The Exact AI Skills This Solo Founder Uses to Build 5 Apps at Once | Josh Pigford Peter Yang Josh Pigford demonstrates his autonomous AI stack—combining Conductor-powered 4-step “/build” with Opus, a GPT-3.5 “/adversarial-code-review,” a “/but-for-real” error checker, and a “/learnings” updater of CLAUDE.md—to solo-build and launch five AI products in parallel.
“Getting Claude to Actually Read Your CLAUDE.md - Shows how to use conditional XML blocks inside CLAUDE.md to ensure Claude attends to the right instructions at the right times.”
#11 📝 HumanLayer Blog Getting Claude to Actually Read Your CLAUDE.md - Shows how to use conditional XML blocks inside CLAUDE.md to ensure Claude attends to the right instructions at the right times. These patterns help make agent behavior more reliable by prioritizing context and instructions.
“Bolt @boltdotnew added support for Claude.md imports, allowing agents to ingest project context instantly without manual setup.”
Bolt @boltdotnew added support for Claude.md imports, allowing agents to ingest project context instantly without manual setup. See update .
Related
Anthropic’s coding product/blog referenced in a customer story about Cognition’s use of Claude Fable 5. For AI PMs, it highlights enterprise coding adoption narratives.
Anthropic’s assistant and coding tool, discussed here in both the Reflection dashboard and a physical-AI deployment at UST. The newsletter highlights its usage analytics, workflow suggestions, and enterprise integration.
An AI builder/commentator mentioned twice in the newsletter, including launching a local daemon for agents. He is also listed as a secondary source on GPT-5.6 coverage.
Writer/observer cited for reframing agent building as a stack of LLM primitives and persistent memory.
Reusable behavior modules or instructions for guiding AI agents. The newsletter mentions skills as one of the steering mechanisms for Claude Code and other agents.
Specialized subordinate agents used to break down and orchestrate tasks. The newsletter mentions them as part of Claude Code steering controls.
A company/platform for AI coding collaboration and SDLC workflows. It is presented as a general-availability launch with workspaces, agents, approvals, and visibility controls.
A collaborative coding environment with live multiplayer, real-time typing, and shared chat history. Relevant to AI PMs building multi-user AI creation tools.
Stay updated on Claude.md
Get curated AI PM insights delivered daily — covering this and 1,000+ other sources.
Subscribe Free