A practical, reusable library for building and operating AI-assisted engineering workflows. The repository brings together role definitions, engineering rules, focused skills, safety gates, and MCP/API connectors that developers select and copy into their own repositories.
Important
These resources are implementation aids, not a substitute for engineering review. Validate every workflow against your environment, security policy, data-handling requirements, and approval boundaries before using it in production.
| Collection | Purpose | Best for |
|---|---|---|
| Daily AI Engineering Kit | Modular gates, guards, workflows, hooks, schemas, and verification assets. | Adding a focused engineering control to an agent workflow. |
| Daily AI Engineering Security - Performance - Thinking | Deep controls for agent safety, performance, context management, permissions, and execution integrity. | Hardening autonomous or long-running agent systems. |
| Daily AI Role | Role packages for engineering, product, operations, architecture, design, and go-to-market disciplines. | Giving an agent a bounded mission, operating model, and expected outputs. |
| MCP-API | Provider-scoped connectors for Discord, GitHub, Jira, Linear, Notion, Slack, Stripe, and Telegram. | Integrating external services through narrow, reviewable capabilities. |
| Rules | Technology- and role-specific operating constraints. | Establishing non-negotiable quality and safety expectations. |
| Skills | Focused procedures for common engineering tasks. | Supplying repeatable, task-level guidance without adopting a full role package. |
Role defines ownership, scope, and expected outcomes
+
Rules establish mandatory constraints
+
Skills provide task-specific procedures
+
Kits add workflows, gates, evidence, and verification
+
MCP/API exposes narrowly scoped external capabilities
Each collection is intentionally modular. This repository is not one application to install or enable all at once. Start with the smallest unit that satisfies the task, copy it into the target repository, and add controls only when the risk or operating environment requires them.
- Choose the collection that matches your goal from the table above.
- For a rule or skill, copy the selected Markdown file; copy its discipline index only when useful for navigation.
- For a role, engineering kit, guard, or connector, copy the complete child package directory so its local schemas, examples, scripts, and templates remain together.
- Read the copied package's
README.md; its package-local prerequisites and commands are authoritative for that selection. - Adapt paths and integration points to the target repository, then run only that package's documented validation.
For example, a .NET backend workflow can combine the .NET Backend Developer role, .NET rules, and .NET skills. A higher-risk change can then add a relevant gate from the engineering kit.
See the adoption guide for selection boundaries, sparse checkout, suggested destinations, dependency rules, and an update checklist. Consumers do not need the root Node.js dependencies or repository-wide audit commands.
For recommended Role, Rule, Skill, gate, and MCP combinations by outcome, use the composition guide. It calls out where a role package is intentionally the only authoritative starting point.
Use the starter packs when you want a small reviewed selection instead of browsing every collection:
- ASP.NET Core API change
- React user-facing feature
- Tool-using AI agent
- Application security review
- Reliability and incident readiness
- Data pipeline change
- Product discovery to acceptance
- Documentation change
Starter packs are navigation recipes, not installable bundles. Remove anything that does not apply to the target task.
- Composable: packages can be adopted independently.
- Evidence-driven: completion should be supported by reproducible checks.
- Least privilege: external actions should expose the narrowest practical capability.
- Human-governed: destructive, sensitive, or externally visible actions require explicit approval.
- Tool-neutral: guidance should remain useful across compatible agent runtimes.
- Production-aware: security, reliability, observability, rollback, and operational handoff are first-class concerns.
Contributions that improve clarity, correctness, safety, coverage, or verification are welcome. Read CONTRIBUTING.md before proposing a change and follow the Code of Conduct in all project spaces.
Please report suspected vulnerabilities through the private process described in SECURITY.md. For usage questions and general help, see SUPPORT.md.
- Documentation hub
- Getting started
- Adoption guide
- Composition guide
- Starter packs
- Content quality standard
- Trust and provenance
- Frequently asked questions
- Repository structure
- Documentation style guide
- Maintenance guide
- Governance
- Roadmap
- Maintainers
- Changelog
- Contributing guide
- Code of Conduct
- Security policy
- Support guide
The following commands maintain this source library; they are not installation steps for copied content. With Node.js 22+, maintainers can install the locked audit dependencies and run the structural audit from the repository root:
npm ci
npm run audit
npm run audit:standaloneThe repository audit checks collection and package READMEs, Rules/Skills indexes, relative Markdown links, JSON/YAML syntax, Python syntax, and required MCP connector files. The standalone audit additionally checks supported copy boundaries, package-local references and dependency declarations, script discoverability, rule/skill contracts, and schema examples. Use npm run audit:strict and npm run audit:standalone:strict to treat documentation maturity findings as blocking. Executable package tests must still be run with their documented runtime.
When changing an MCP connector itself, work inside that connector directory and run only the build and test commands documented in its package README. Do not use a repository-wide build as a substitute for provider-specific validation.
Distributed under the MIT License. Preserve the applicable license notice when copying or redistributing selected content.