Packmind Engineering Playbook Platform

Packmind's design system, themed as "The Curator's Workshop," focuses on a calm, warm-dark aesthetic. It is designed to be low-chrome and content-led, avoiding common enterprise-B2B or AI-startup visual tropes (like purple gradients or glassmorphism).

Key Visual Principles:

• Warm-Dark Surfaces: Uses a beige-based neutral scale instead of cool grays.
• Single Brand Accent: Uses periwinkle blue (blue-200) sparingly for primary affordances.
• Single Typeface: Uses the 'Borna' font family for all hierarchy levels.
• Flat Tonal Depth: Depth is achieved through four tonal background steps rather than shadows.
• Low-Chrome Components: Minimalist borders and deliberate fills to keep the interface out of the way of the content.

Packmind is a developer tool designed to centralize and distribute an engineering playbook to AI coding agents. It solves the problem of fragmented engineering standards (architecture rules, naming conventions, patterns, and best practices) by allowing teams to define these rules once and automatically generating the specific instruction files required by different AI tools.

Packmind targets compatibility with various AI assistant formats, including:

• GitHub Copilot: .github/copilot-instructions.md, chat modes, and reusable prompts.
• Claude: CLAUDE.md, commands, and skills.
• Cursor: .cursor/rules/*.mdc files.

By using Packmind, you ensure that AI agents follow your team's specific coding standards and patterns without manual copy-pasting across repositories or tool-specific configuration files.

Packmind's linter uses AI to generate deterministic JavaScript detection programs that automatically identify violations of your team's coding rules. This feature is an Enterprise capability.

Prerequisites:

• Enterprise License: This functionality is exclusive to the Enterprise version.
• LLM Connection: Requires a Large Language Model connection. While enabled by default on Packmind Cloud, self-hosted installations require an OpenAI API key configuration.

The @packmind/feature-flags package is a pure, browser-safe shared registry used to manage Packmind's feature flags. It contains the registry of feature keys and email-domain rules, along with the isFeatureFlagEnabled decision function.

Because it is consumed by both the frontend (@packmind/ui) and the backend, it ensures that feature flag decisions remain consistent across the entire stack.

Important Constraint: This package is tagged env:shared and must remain browser-safe. Do not add Node.js-only dependencies like @packmind/node-utils or TypeORM to this package.

Packmind is an engineering playbook management system designed for AI-native development. It solves the problem of fragmented AI instructions (e.g., .github/copilot-instructions.md, CLAUDE.md, .cursor/rules/*.mdc) by centralizing your team's standards, architecture rules, naming conventions, and best practices in one place.

Instead of manually copy-pasting instructions across different repositories and AI tools, Packmind:

1. Centralizes your playbook once.
2. Distributes it everywhere by generating the specific instruction files required by different AI agents (Claude Code, GitHub Copilot, Cursor, etc.).
3. Governs the content through controlled editing and approval workflows, ensuring that context files are updated via proposals from the project codebase rather than unmanaged PRs.

A marketplace is a Git repository that hosts a marketplace.json descriptor, which acts as a catalogue for Claude Code to discover and install plugins.

Instead of manually committing instruction files into every project, you publish a Packmind package once to a marketplace. Packmind then automatically renders the package's skills and commands into a plugin and manages the updates via a Packmind sync pull request on your marketplace repository.

Key capabilities:

• Publish packages as managed plugins for your entire organization.
• Track the lifecycle of every publish attempt.
• Monitor marketplace health and plugin adoption.
• Retire plugins using a governed pull-request flow.

When recording, you must use a consistent tool prefix for all browser interactions. Mixing different Playwright MCP prefixes will cause "Browser is already in use" errors because they operate in different browser contexts.

Correct Prefix: Use mcp__playwright__browser_* for all commands, including:

• browser_navigate
• browser_click
• browser_type
• browser_snapshot
• browser_wait_for
• browser_evaluate
• browser_close
• browser_start_video

Incorrect Prefix: Do NOT use mcp__plugin_playwright_playwright__browser_* during a recording session.

When running Packmind on your own infrastructure, you have complete local control over your data. No data is sent to Packmind's cloud services, and Packmind has no access to your database or operational data. This allows for air-gapped deployments where the application runs entirely offline within your local network.

Key Security Features

• Complete Local Control: All data, databases, and backups remain within your infrastructure.
• No Data Sharing: Packmind cannot see your activities or access your data.
• Air-Gapped Support: Full functionality is available without an internet connection using Docker Compose or Kubernetes.
• Data Retention: You define all retention and deletion policies; Packmind performs no automatic deletions.