SaaS AI Skills Collection

Independent, portable AI skills for planning, building, reviewing, securing, testing, and operating modern SaaS applications. Each skill provides a focused, evidence-based workflow without requiring a shared runtime or another installed skill.

Start here

Browse the skill catalog and choose the primary outcome you need.
Install the collection, one skill, or a selected set using the installation guide.
Give your compatible coding agent the relevant request and project artifacts.
Read FAQ, troubleshooting, or support if you need help.

# Install the collection
npx skills add noorkhalel/saas-ai-skills

# Install one skill
npx skills add noorkhalel/saas-ai-skills --skill debugging

# Install a focused workflow
npx skills add noorkhalel/saas-ai-skills \
  --skill codebase-understanding \
  --skill security-audit \
  --skill test-generation

The folder name is the installation identifier. Check npx skills add --help if your installed Skills CLI differs.

Choose a skill

If you need to...	Use
Understand an unfamiliar repository or trace a feature	Codebase Understanding
Plan a new system or material future-state change	Architecture Planning
Review current boundaries and modernization options	Clean Architecture Review
Review a change or implementation broadly	Code Review
Diagnose an active failure	Debugging
Produce a post-incident systemic analysis	Root Cause Analysis
Design an API, schema, tests, or safe refactor	API Design Review, Database Design, Test Generation, or Refactoring Code
Audit security, performance, or dependencies	Security Audit, Performance Optimization, or Dependency Analysis
Evaluate SOLID risks or a design-pattern choice	SOLID Review or Design Pattern Advisor

See SKILLS.md for all 15 skills, their categories, descriptions, and exact installation identifiers. The routing matrix explains close calls.

What makes these skills dependable

Evidence first. Findings distinguish verified facts, inferences, assumptions, and unknowns.
Clear routing. The requested deliverable selects the skill; adjacent concerns do not automatically expand scope.
Safe standalone packaging. Every installed skill carries the policy subset, references, and optional workflow contract it needs.
Context discipline. References are loaded only when they can change the next decision.
Validated behavior. CI checks routing, generated catalogs, packaged resources, workflow persistence, redaction, prompt-injection resistance, report schemas, and deterministic evaluation fixtures.

Read Architecture for the system design and Quality System for the validation model.

Optional cross-skill workflow

Skills remain independent. When a project opts in by creating .ai-workflow/ or requesting persistent output, they can leave compact, re-verifiable handoffs for later skills:

.ai-workflow/
  state.json                 # lightweight run metadata
  artifacts/<skill-name>.md  # detailed specialist report
  handoffs/<skill-name>.json # concise, topic-filtered summary

Handoffs are leads, not facts: the receiving skill verifies important claims in the project before relying on them. Missing or malformed workflow files never block a normal standalone run. See the workflow contract.

Compatibility

Surface	Support statement
Skills CLI	Uses repository installation and repeated `--skill <folder>` selectors. Verify your local CLI with `npx skills add --help`.
Folder-based loaders	Supported when the loader can read `SKILL.md` and relative package files.
Single-rules-file tools	Use `SKILL.md` and inline only the needed relative references.
Optional integrations	Git, filesystem, database, browser, observability, and MCP tools improve evidence collection but are never required for the skill to produce a bounded result.

Documentation

Topic	Read
Installation and compatibility	INSTALL.md
Catalog and selection	SKILLS.md and shared/routing-matrix.md
Architecture, policies, persistence, and handoffs	ARCHITECTURE.md
Evaluations, validation, and CI	QUALITY_SYSTEM.md
Contributing and development	CONTRIBUTING.md and DEVELOPMENT.md
Releases and changes	RELEASE.md, RELEASE_NOTES.md, and CHANGELOG.md
Security and community	SECURITY.md, SUPPORT.md, and CODE_OF_CONDUCT.md

Release status

Version 1.1.0 introduces the Base Framework, deterministic routing and evaluation layers, standalone packaging checks, optional workflow persistence, and release-quality validation. See release notes for upgrade information and known limitations.

Contributing and security

Contributions are welcome. Start with CONTRIBUTING.md, use the provided issue and pull-request templates, and run the documented validation before opening a PR. Please report security concerns through the process in SECURITY.md, not in a public issue.

This project is licensed under the MIT License.

Saas AI Skills

README