Observed Signal · May 4, 2026 · Technical Release · Source: DEV Community · Impact: 3/5 · Sentiment: Positive
STRATEGY.md as Code: LLM Contract for Agents
The article proposes a pattern called “STRATEGY.md as code”: writing strategy as a machine-loadable contract file that LLM agents must load at the start of each session so recommendations respect local business rules. The author (Yoshinaga, founder of mureo) describes six core sections for such a file (Persona, USP, Brand voice, Goals, Operation mode, Constraints) and gives ad-ops examples where constraints prevent repeated mistakes. The piece details mureo v0.8.0 (released 2026-05-02), which ships five MCP tools that expose STRATEGY.md and STATE.json to agents (read/write) so any MCP-capable host can load the contract. The pattern generalizes beyond marketing to SRE incident response, investment theses, and threat models; the author stresses the contract is only as good as the people who write and maintain it.
Introduces an operational pattern that grounds LLM agents with account-specific constraints and an accompanying mureo release (MCP tools) that enables cross-host loading; relevant to ad-ops/MarTech implementations but not a major platform policy change.
Track Meta Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Article authored by Yoshinaga (founder, mureo) and published on 2026-05-04.
- Proposed pattern: write strategy as a machine-loadable contract file ('STRATEGY.md as code') containing six sections: Persona, USP, Brand voice, Goals, Operation mode, Constraints.
- mureo v0.8.0 (released 2026-05-02) shipped five MCP tools: mureo_strategy_get, mureo_strategy_set, mureo_state_get, mureo_state_action_log_append, mureo_state_upsert_campaign.
- The mureo project and its strategy-format documentation are available in a GitHub repository (github.com/logly/mureo).
- Pattern is positioned as applicable beyond marketing — e.g., INCIDENT_RESPONSE.md, INVESTMENT_THESIS.md, THREAT_MODEL.md — wherever local rules diverge from generic LLM training data.
Connected Companies & Entities
5 Entities mappedOntology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
AI Instruction Split: AGENTS.md, SKILL.md, DESIGN.md
The article describes a growing three-layer standard for instructing AI agents: AGENTS.md for overall agent behavior and boundaries, SKILL.md for reusable task procedures (used by Anthropic's Claude Skills and the Agent Skills standard), and DESIGN.md — a Google Labs design-spec format released in April 2026 that combines machine-readable design tokens (YAML) with human-readable intent and ships with a CLI validator (npx @google/design.md lint). The author argues these formats separate verifiable rules (tokens, audits, structural checks) from judgment-based guidance (tone, stance), and situates the split alongside Spec-Driven Development (SDD) workflows (Kiro, GitHub Spec Kit). The three-layer approach is presented as complementary to SDD and intended for incremental adoption where verification adds value.
Write MCP Server Contracts for LLM Clients
The article explains that the Model Context Protocol (MCP) effectively turns internal tools into APIs whose primary consumer is a language model (an agent) rather than a human developer. Because agents read tool schemas at runtime and act automatically, server-side contracts must be precise: express constraints in JSON Schema, provide structured outputs and stable field names, supply an error taxonomy with actionable codes and field-level validation details, and annotate side effects (read-only, idempotent, destructive) so agents can decide when to retry or escalate. The piece shows patterns (idempotency keys, dry_run + confirm_token flows, append-only error codes) and offers a sample refund tool contract to illustrate versioning and breaking-change risks for agent consumers.
Viral CLAUDE.md File Teaches Branding Agencies AI Discipline
A markdown file containing four behavioral rules for AI coding agents, based on observations by AI researcher Andrej Karpathy, went viral on GitHub, gaining 91,000 stars. The article analyzes why the file resonated with developers and draws parallels between AI coding discipline and branding workflows at California agencies. It argues that the viral success highlights a widespread need for structured behavioral contracts with AI tools, applicable beyond software development to creative and marketing operations. The author, Mollie Nelson, discusses principles such as 'Think Before You Code', 'Simplicity First', 'Surgical Changes Only', and 'Goal-Driven Execution', and suggests that branding agencies can adopt similar 'BRAND.md' files to set tone, scope, and constraints for AI-generated content. The article includes caveats about the limitations of such instruction files and offers practical steps for brands and agencies to improve AI collaboration.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.