Skip to content
/
Tier A A4 Markdown scaffold

Jeffallan/claude-skills

jeffallan-claude-skills

Primitive shape
No installable primitives
00

Summary

Jeffallan/claude-skills — Summary

jeffallan/claude-skills is a 66-skill Claude Code plugin focused on full-stack development with a clean taxonomy of 12 technology categories. The feature-forge skill conducts structured requirements workshops producing EARS-format specifications, user stories, and acceptance criteria — the only EARS-format spec skill in this batch. A common-ground command surfaces and validates Claude's hidden assumptions about project context before implementation. Nine workflow commands manage epics from discovery through retrospectives, integrating with Jira and Confluence via Atlassian MCP. Multi-skill workflows are explicitly documented: "Feature Development: Feature Forge → Architecture Designer → Fullstack Guardian → Test Master → DevOps Engineer." The spec-miner skill extracts existing behavior from codebases to create specs retrospectively. the-fool skill provides adversarial review. Full documentation website at jeffallan.github.io/claude-skills. Compared to seeds, jeffallan occupies the skills-only behavioral framework archetype (like superpowers) but is organized as a developer marketplace rather than an "Iron Law" behavioral enforcer, and uniquely ships EARS-format requirement specification and Jira/Confluence integration as first-class features.

01

Overview

Jeffallan/claude-skills — Overview

Origin

Created by Jeff Smolinski / Jeffallan (Principal Consultant at Synergetic Solutions). 9,353 stars, last push 2026-05-20. v0.4.15. Mentioned in Awesome Claude Code.

Philosophy

From README:

"66 specialized skills across 12 categories covering languages, backend/frontend frameworks, infrastructure, APIs, testing, DevOps, security, data/ML, and platform specialists."

Focus on context-aware activation and multi-skill workflows:

"Skills activate automatically based on your request: 'Implement JWT authentication in my NestJS API' → Activates: NestJS Expert → Loads: references/authentication.md"

Multi-Skill Workflow Documentation

Explicit workflow chains:

  • Feature Development: Feature Forge → Architecture Designer → Fullstack Guardian → Test Master → DevOps Engineer
  • Bug Investigation: Debugging Wizard → Framework Expert → Test Master → Code Reviewer
  • Security Hardening: Secure Code Guardian → Security Reviewer → Test Master

Common Ground Innovation

/common-ground — surfaces Claude's hidden assumptions about project context:

"Claude often operates on assumptions about project context, technology choices, coding standards, and user preferences. This command surfaces those assumptions for explicit user validation, preventing misaligned work based on incorrect premises."

Atlassian Integration

Nine workflow commands require Atlassian MCP for Jira/Confluence integration — the most enterprise-CRM-integrated skill collection in this batch.

Documentation

Full documentation website: jeffallan.github.io/claude-skills

  • Skills Guide with decision trees
  • Workflow Commands Reference
  • Common Ground Guide
  • Atlassian MCP Setup Guide
02

Architecture

Jeffallan/claude-skills — Architecture

Distribution

Claude Code plugin marketplace + Astro documentation site.

/plugin marketplace add jeffallan/claude-skills
/plugin install fullstack-dev-skills@jeffallan

Directory Structure

skills/                      # 66 skill directories
  angular-architect/
  api-designer/
  architecture-designer/
  atlassian-mcp/
  chaos-engineer/
  feature-forge/
  fullstack-guardian/
  spec-miner/
  test-master/
  the-fool/
  typescript-pro/
  ...
commands/                    # Slash commands
  common-ground/             # Assumption surfacing command
    COMMAND.md
    common-ground.yaml
    references/
  intake/                    # Project intake command
  project/                   # Project management commands
  workflow-manifest.yaml
.claude-plugin/              # Plugin registry
  marketplace.json
  plugin.json
.astro/                      # Documentation site source
docs/
  COMMON_GROUND.md
  WORKFLOW_COMMANDS.md
  ATLASSIAN_MCP_SETUP.md
  local_skill_development.md
site/                        # Built documentation site
specs/                       # Spec templates

Required Runtime

  • Node.js (for Atlassian MCP server)
  • Atlassian MCP server (for workflow commands with Jira/Confluence)
  • Python (ruff.toml present — Python linting configured)

Target AI Tools

Claude Code only.

Documentation

Astro-powered documentation site at jeffallan.github.io/claude-skills.

03

Components

Jeffallan/claude-skills — Components

Totals

Type Count
Skills 66
Commands 9 (workflow commands) + 3 (common-ground, intake, project)
Subagents 0
Hooks 0

Skills by Category (66 total)

Languages (8)

  • typescript-pro, javascript-pro, python-pro, golang-pro
  • rust-engineer, java-architect, kotlin-specialist, cpp-pro

Backend Frameworks (7)

  • nestjs-expert, spring-boot-engineer, fastapi-expert
  • django-expert, rails-expert, laravel-specialist, php-pro

Frontend (6)

  • react-expert, vue-expert, vue-expert-js, angular-architect
  • nextjs-developer, react-native-expert, flutter-expert

Infrastructure (7)

  • kubernetes-specialist, terraform-engineer, devops-engineer
  • sre-engineer, cloud-architect, monitoring-expert, microservices-architect

APIs & Data (6)

  • api-designer, graphql-architect, postgres-pro, sql-pro
  • database-optimizer, websocket-engineer

Testing (3)

  • test-master, playwright-expert, chaos-engineer

Security (2)

  • secure-code-guardian, security-reviewer

Data/ML (5)

  • ml-pipeline, spark-engineer, pandas-pro, rag-architect, fine-tuning-expert
  • prompt-engineer

DevOps (2)

  • devops-engineer, sre-engineer

Platform Specialists (4)

  • shopify-expert, salesforce-developer, wordpress-pro, mcp-developer

Workflow (4 — these are the most distinctive)

  • feature-forge — EARS-format requirements workshops
  • architecture-designer — system architecture design
  • fullstack-guardian — full-stack implementation oversight
  • spec-miner — extract behavior from existing code as specs

Meta

  • the-fool — adversarial review perspective
  • code-reviewer — standard code review
  • code-documenter — documentation generation
  • debugging-wizard — systematic debugging
  • legacy-modernizer — legacy code modernization
  • cli-developer — CLI tool development

Commands (12)

common-ground (3 modes)

  • Default: Surface & Adjust two-phase interactive flow
  • --list: Read-only view of tracked assumptions
  • --check: Quick validation of current assumptions
  • --graph: Mermaid diagram of reasoning structure

Workflow Commands (9, Atlassian MCP required)

Epic lifecycle from discovery through retrospective — integrate with Jira and Confluence.

intake + project Commands

Project intake and management commands.

Feature-Forge Skill (Notable)

Implements EARS-format (Easy Approach to Requirements Syntax) requirement writing:

  • PM Hat + Dev Hat dual perspective
  • Structured AskUserQuestion-based requirements gathering
  • EARS functional requirements + acceptance criteria (Given/When/Then)
  • Pre-discovery subagents for multi-domain features

References:

  • references/ears-syntax.md — EARS format spec
  • references/interview-questions.md — requirements gathering guide
  • references/specification-template.md — spec output template
  • references/acceptance-criteria.md — Given/When/Then format
  • references/pre-discovery-subagents.md — multi-domain discovery
05

Prompts

Jeffallan/claude-skills — Prompt Excerpts

Excerpt 1: feature-forge SKILL.md

Technique: Dual-perspective role (PM Hat + Dev Hat) + EARS format requirement writing

---
name: feature-forge
description: Conducts structured requirements workshops to produce feature specifications, user stories, EARS-format functional requirements, acceptance criteria, and implementation checklists. Use when defining new features, gathering requirements, or writing specifications. Invoke for feature definition, requirements gathering, user stories, EARS format specs, PRDs, acceptance criteria, or requirement matrices.
metadata:
  version: "1.1.0"
  domain: workflow
  role: specialist
  related-skills: fullstack-guardian, spec-miner, test-master
---

# Feature Forge

Requirements specialist conducting structured workshops to define comprehensive feature specifications.

## Role Definition
Operate with two perspectives:
- **PM Hat**: Focused on user value, business goals, success metrics
- **Dev Hat**: Focused on technical feasibility, security, performance, edge cases

## Core Workflow
1. **Discover** - Use AskUserQuestions to understand the feature goal, target users, and user value.
2. **Interview** - Systematic questioning from both PM and Dev perspectives.
3. **Document** - Write EARS-format requirements
4. **Validate** - Use AskUserQuestions to review acceptance criteria with stakeholder
5. **Plan** - Create implementation checklist

## Reference Guide
| Topic | Reference | Load When |
|---|---|---|
| EARS Syntax | `references/ears-syntax.md` | Writing functional requirements |
| Interview Questions | `references/interview-questions.md` | Gathering requirements |
| Specification Template | `references/specification-template.md` | Writing final spec |
| Pre-Discovery Subagents | `references/pre-discovery-subagents.md` | Multi-domain features |

Analysis: related-skills in metadata establishes explicit skill chain dependency (feature-forge → fullstack-guardian → test-master). Reference guide table with load-when conditions implements progressive disclosure — full reference material not loaded into context unless needed. Dual PM/Dev perspective pattern avoids single-perspective bias.

Excerpt 2: common-ground COMMAND.md

Technique: Assumption taxonomy + project identification via git + Mermaid diagram mode

---
name: common-ground
description: Surface and validate Claude's hidden assumptions about the project for user confirmation
argument-hint: "[--list] [--check] [--graph]"
---

## Argument Parsing
| Flag | Mode | Description |
|---|---|---|
| (none) | Default | Surface & Adjust two-phase interactive flow |
| `--list` | List | Read-only view of all tracked assumptions |
| `--check` | Check | Quick validation of current assumptions |
| `--graph` | Graph | Generate mermaid diagram of reasoning structure |

## Project Identification
1. Try git remote: `git remote get-url origin 2>/dev/null`
   If found, use URL as project identifier
2. Fallback to current working directory path

Store as {project_id} for file operations.

Analysis: Git-remote-based project identification is a reliable non-invasive fingerprinting approach. The --graph mode generating Mermaid diagrams of reasoning structure is unique — no other skill in this batch generates visual dependency graphs of assumptions.

09

Uniqueness

Jeffallan/claude-skills — Uniqueness & Differentiation

Distinguishing Characteristics

1. EARS-Format Requirements (feature-forge)

The feature-forge skill mandates use of EARS (Easy Approach to Requirements Syntax) for functional requirement writing. EARS is a structured natural-language pattern language developed in aerospace/defense that constrains requirement sentences to a small set of templates (WHEN <event> THE <system> SHALL <response>, etc.). No other catalog in this batch applies a formal requirements notation.

2. Dual-Perspective Role Pattern (PM Hat + Dev Hat)

feature-forge operates with two simultaneous personas: a Product Manager perspective (user value, business goals, metrics) and a Developer perspective (technical feasibility, security, performance, edge cases). This structured perspective rotation is a systematic defense against single-viewpoint bias in requirements. The pattern is explicit in the SKILL.md role definition, not emergent.

3. Adversarial Review Skill (the-fool)

An explicit skill named the-fool provides adversarial critique. This is a dedicated anti-yes-man persona distinct from any normal code reviewer. The skill's role is to find failure modes, challenge assumptions, and surface risks the main workflow missed. Naming it after a Tarot archetype (the Fool = beginner's mind / chaos-revealing) signals its purpose.

4. Retrospective Specification (spec-miner)

The spec-miner skill extracts behavioral specs from existing code — the inverse of the conventional spec-first workflow. This enables generating specifications for legacy systems that never had formal specs, closing the gap between deployed behavior and documented intent.

5. Git-Remote-Based Project Fingerprinting (common-ground)

The common-ground command identifies the current project via git remote get-url origin, using the remote URL as a stable cross-session key for storing the assumption registry. This is a low-friction, non-invasive approach: no config files, no IDs, no manual project naming. Fallback to cwd path when no remote exists.

6. Assumption Graph Visualization (--graph mode)

common-ground --graph generates a Mermaid diagram of the reasoning dependency structure between assumptions. This is unique in the batch: no other catalog generates a visual graph of Claude's epistemic state (what it assumes, which assumptions depend on others).

7. Atlassian MCP Integration (9-phase enterprise workflow)

The 9-command Atlassian workflow pipeline covers the full lifecycle from epic discovery to retrospective, creating real Jira epics/stories/tasks and Confluence documentation as first-class artifacts. This is the only catalog in the batch with this depth of external project management integration.

8. Progressive Disclosure Reference Tables

Each skill's reference table includes a Load When condition column, making context loading explicit and conditional. This moves progressive disclosure from an implicit convention (load only what you need) to a documented, machine-readable specification (load this reference when this condition is met).

Comparison to Batch Peers

Dimension jeffallan wshobson/agents daymade qdhenry
Formal requirements notation EARS None None None
Adversarial review skill Yes (the-fool) No No No
Assumption management common-ground command None None None
Retro-spec extraction Yes (spec-miner) No No No
External PM integration Atlassian MCP (9 cmds) None None None
Documentation site Yes (Astro) No No No

Limitations

  • No automated skill chaining — related-skills metadata is documentation, not orchestration.
  • Atlassian workflow commands require external MCP server setup; without it, 9 of 12 commands degrade significantly.
  • No cross-session memory beyond the common-ground ground file, which is manually loaded.
  • Single-tool target (Claude Code only); no multi-runtime export unlike voltagent-subagents.
04

Workflow

Jeffallan/claude-skills — Workflow

Multi-Skill Workflow Chains

Feature Development

Feature Forge → Architecture Designer → Fullstack Guardian → Test Master → DevOps Engineer

Bug Investigation

Debugging Wizard → Framework Expert → Test Master → Code Reviewer

Security Hardening

Secure Code Guardian → Security Reviewer → Test Master

Common Ground Workflow (Pre-Implementation)

  1. Run /common-ground — surfaces assumptions in two phases:
    • Surface: Claude identifies its hidden assumptions (tech stack, coding standards, preferences)
    • Adjust: User validates/corrects assumptions
  2. Save assumption set to ground file (project-specific storage)
  3. Future sessions load ground file on start

Feature Development With feature-forge

  1. Activate feature-forge skill
  2. Discover phase: AskUserQuestion for goal, target users, user value
  3. Interview phase: PM Hat + Dev Hat questions
  4. Document phase: Write EARS-format requirements
  5. Validate phase: AskUserQuestion to review acceptance criteria
  6. Plan phase: Create implementation checklist with TODO items

Atlassian Workflow Integration

9 workflow commands require Atlassian MCP:

  1. Epic discovery (Confluence doc creation)
  2. Feature refinement (Jira story creation)
  3. Implementation sprint (Jira task tracking)
  4. Code review gate
  5. QA testing
  6. Deployment
  7. Documentation update (Confluence)
  8. Retrospective (Jira)
  9. Metrics collection

Phase-to-Artifact Map

Phase Artifact
common-ground Project assumption registry file
feature-forge Discover Requirements draft
feature-forge Document EARS-format spec + acceptance criteria
feature-forge Plan Implementation checklist
Atlassian workflow Jira epics/stories/tasks + Confluence docs
06

Memory Context

Jeffallan/claude-skills — Memory & Context

Session Memory

No persistent session memory built into the skill system. Skills do not write state between invocations; each Claude Code session starts with a blank context window.

Cross-Session Persistence: common-ground Command

The common-ground command (commands/common-ground/COMMAND.md) is the only mechanism for cross-session state:

  1. Surface phase: Claude identifies its hidden assumptions about the project (tech stack, coding conventions, preferences, environmental constraints).
  2. Adjust phase: User validates and corrects those assumptions.
  3. Persist: Validated assumption set is written to a ground file keyed by project identifier.

Project identification logic:

  • Primary: git remote get-url origin 2>/dev/null → URL becomes the key.
  • Fallback: current working directory path.

The assumption registry file is a Markdown file stored locally. Future sessions are expected to load the ground file at session start. The system is manual — there is no automatic injection of the ground file into context; users must reference it.

Progressive Disclosure for Context Management

Skills cap inline content and offload details to references:

  • Each skill's SKILL.md is kept brief.
  • Reference tables include a Load When column specifying the condition under which the reference file should be read into context.

Example from feature-forge:

Reference Load When
references/ears-syntax.md Writing functional requirements
references/interview-questions.md Gathering requirements
references/specification-template.md Writing final spec
references/pre-discovery-subagents.md Multi-domain features

This prevents all reference material from polluting context when not required.

Skill frontmatter includes a related-skills list:

metadata:
  related-skills: fullstack-guardian, spec-miner, test-master

This is declarative documentation for a skill chain, not an automated context-loading mechanism. The user or developer must manually activate next-chain skills.

No Audit Log

The system writes no structured log of past tool invocations, skill activations, or requirement decisions. Ground files are the only cross-session artifact.

Memory Summary

Mechanism Scope Persistence Automation
common-ground ground file Project-level assumptions Cross-session Manual
Progressive disclosure references Skill body context Within session Conditional on trigger
related-skills metadata Skill chain hints None (docs only) None
07

Orchestration

Jeffallan/claude-skills — Orchestration

Orchestration Model

Single-agent, sequential. There is no automated multi-agent dispatch. The documented skill chains (e.g., Feature Forge → Architecture Designer → Fullstack Guardian → Test Master → DevOps Engineer) are workflow guidance, not automated pipelines.

Skill Chains (Manual)

Documented in each skill's related-skills metadata and in README workflow documentation:

Feature Development

Feature Forge → Architecture Designer → Fullstack Guardian → Test Master → DevOps Engineer

Bug Investigation

Debugging Wizard → Framework Expert → Test Master → Code Reviewer

Security Hardening

Secure Code Guardian → Security Reviewer → Test Master

Each step in the chain requires the user to explicitly activate the next skill. There is no automatic handoff.

Atlassian Workflow Commands (9-step enterprise pipeline)

The workflow commands in commands/workflow/ form the most structured orchestration in the system. These require an external Atlassian MCP server (Jira + Confluence):

  1. Epic discovery → Confluence doc creation
  2. Feature refinement → Jira story creation
  3. Implementation sprint → Jira task tracking
  4. Code review gate
  5. QA testing
  6. Deployment
  7. Documentation update → Confluence
  8. Retrospective → Jira
  9. Metrics collection

These commands are sequential and stateful (each step creates artifacts consumed by the next), but execution is user-initiated per step, not automated.

feature-forge Internal Workflow

Within a single skill activation, feature-forge has a structured 5-phase loop:

  1. Discover — AskUserQuestions for goal, users, value
  2. Interview — PM Hat + Dev Hat questions
  3. Document — write EARS-format requirements
  4. Validate — AskUserQuestions on acceptance criteria
  5. Plan — implementation checklist

This is intra-skill orchestration; the skill drives the conversation flow via AskUserQuestion tool calls.

common-ground Two-Phase Flow

  1. Surface: Claude lists hidden assumptions
  2. Adjust: User confirms or corrects
  3. Persist: Write ground file

Also intra-command, not multi-agent.

No Subagent Spawning

The repo contains no use of Claude Code's Task tool for spawning subagents. All work is done within the single active skill/command invocation.

Parallelism

None documented. Skills are designed for sequential single-agent execution.

External Integration (MCP)

The Atlassian workflow commands require an external Atlassian MCP server, configured separately by the user. The skills themselves contain the workflow logic; MCP provides the Jira/Confluence API surface. This is the only external service dependency.

Orchestration Summary

Pattern Used Notes
Multi-agent subagent spawning No Single-agent throughout
Automated skill chaining No Related-skills is documentation only
Intra-skill conversation loop Yes feature-forge 5-phase, common-ground 2-phase
External MCP orchestration Yes (conditional) Atlassian workflow requires MCP server
Hooks / lifecycle events No No PreToolUse/PostToolUse hooks
08

Ui Cli Surface

Jeffallan/claude-skills — UI / CLI Surface

Distribution

Published to the Claude Code plugin marketplace as:

  • Plugin: fullstack-dev-skills@jeffallan

Installation via Claude Code:

/plugins install fullstack-dev-skills@jeffallan

No separate CLI binary or npm package.

Documentation Site

An Astro-based documentation website is maintained at jeffallan.github.io/claude-skills. The site lists all 66 skills across 12 categories with descriptions and usage guidance. This is the primary browsable UI for the catalog.

Skill Invocation Surface

Skills are invoked by natural language trigger within Claude Code. Descriptions in each SKILL.md frontmatter contain trigger phrases — the plugin marketplace uses these to route user requests to the correct skill.

Commands

Three non-workflow commands provide distinct slash-command surfaces:

Command Invocation Modes
common-ground /common-ground --list, --check, --graph
intake /intake unspecified
project /project unspecified

The --graph flag of common-ground produces a Mermaid diagram of the reasoning/assumption dependency structure, rendered in any Mermaid-capable viewer.

Workflow Commands (Atlassian MCP)

Nine workflow commands are surfaced as slash commands requiring an external Atlassian MCP server. These are not available without prior MCP configuration by the user.

No Local Dashboard

No local web server, TUI, or desktop app. Claude Code is the sole interaction surface.

Developer Tooling

The repo includes developer-facing tooling that is not user-facing:

  • ruff.toml — Python linting configuration
  • pyrightconfig.json — Python type checking configuration

These are used for maintaining code quality within Python files bundled with skills (e.g., QA project initialization scripts), not for end-user interaction.

Related frameworks

same archetype · same primary tool · same memory type

BMAD-METHOD ★ 48k
A4 Markdown scaffold

Provides a full agile delivery lifecycle with named expert-persona AI collaborators that elicit the human's best thinking rather…

Agent OS ★ 4.6k
A4 Markdown scaffold

Extracts implicit codebase conventions into token-efficient markdown standards files and injects them selectively into AI agent…

Claude Conductor ★ 367
A4 Markdown scaffold

Gives Claude Code a persistent, cross-linked, auto-analyzed documentation system so it retains codebase context across sessions.

Spec-Driver (Greenfield Spec-Driven Development) ★ 25
A4 Markdown scaffold

Prevents spec rot in AI-assisted development by making implementation changes flow back into evergreen, authoritative specs via…

Anthropic Knowledge Work Plugins ★ 16k
A4 Markdown scaffold

Role-specialized plugin bundles with live MCP connectors that turn Claude into a domain expert for enterprise knowledge workers.

Codex Integration for Claude Code (skill-codex) ★ 1.3k
A4 Markdown scaffold

Single Claude Code skill that handles Codex CLI invocation correctly (stdin blocking, thinking token suppression, session resume)…