Proven AI agent skills for code review, content creation, project management, and more. Built on the Agent Skills open standard and portable across Claude Code, OpenAI Codex, Cursor, GitHub Copilot, and other capable agents.
Slop detection is now a free hosted tool. Slopcheck at tools.synthesiswriting.org/slopcheck/ runs the upgraded synthesis-content-quality and synthesis-fact-checking skills as a web app, with zero data collection and no signup. Same engine that ships with these skills, available without installing anything.
Two major skill upgrades shipped in May 2026. synthesis-content-quality reached v4.0 with model-family fingerprinting across eight LLM families (Claude, GPT, Gemini, Llama, Grok, DeepSeek, Mistral, Qwen), a substance-and-depth section grounded in the Frankfurt-Pennycook-Hicks-Humphries-Slater framework, the compounding-archive principle that retains patterns across the LLM era, and per-family two-axis calibration with an ESL safe-harbor. synthesis-fact-checking reached v2.0 with nine new protocol sections covering nested attribution, paraphrase drift, composite quotes, position-shifting, source-translation drift, URL rot vs hallucination, AI-generated synthetic sources, citation laundering chains, and tool-specific hallucination patterns by LLM family. See CHANGELOG.md for the full v3.0.0 release notes.
One command — installs all skills to every AI agent on your machine:
npx skills add synthesisengineering/synthesis-skills --global --all --copyThis works with Claude Code, OpenAI Codex, Cursor, GitHub Copilot, and 40+ other agents.
curl -fsSL https://raw.githubusercontent.com/synthesisengineering/synthesis-skills/main/install.sh | shOr clone and run directly:
git clone https://github.com/synthesisengineering/synthesis-skills.git
cd synthesis-skills
./install.sh installnpx skills add synthesisengineering/synthesis-skills --global --copy --skill synthesis-pr-review,synthesis-codebase-reviewnpx skills update # Or: ./install.sh update
npx skills remove synthesis-skills # Or: ./install.sh uninstallThe project-management skills use a three-tier memory structure:
CONTEXT.mdfor current working stateREFERENCE.mdfor stable project factssessions/for historical session archives
That structure keeps project memory in the repo, not inside one assistant's chat transcript or tool-native memory. A project can move between Claude Code and Codex, and between synced workstations, because every agent reloads the same durable project files.
All skills are prefixed with synthesis- to prevent namespace collisions with skills from other repositories.
| Skill | Description |
|---|---|
synthesis-codebase-review |
Enterprise-scale codebase audit with tiered review system |
synthesis-code-audit |
10-dimension quality scan of code diffs with scored PASS/WARNING/FAIL verdicts |
synthesis-pr-review |
Delta review methodology with security scanning and AI-analysis verification |
synthesis-review-triage |
PR queue prioritization: scoring, author-response detection, and review routing |
synthesis-code-integration |
Adopt-and-adapt pattern for integrating multi-contributor code with cherry-pick safety |
synthesis-code-planning |
Structured multi-approach evaluation before coding |
synthesis-preflight |
Pre-merge quality gate: tests, types, audit, commit hygiene, go/no-go verdict |
synthesis-implementation-integrity |
Adversarial self-review: verify implementations are genuinely complete before shipping |
| Skill | Description |
|---|---|
synthesis-article-writing |
Two-phase workflow: research/validation then strategic writing |
synthesis-content-distribution |
Strategic content sharing and distribution across platforms with quick-start templates |
synthesis-link-research |
Find authoritative links for people, organizations, and entities |
| Skill | Description |
|---|---|
synthesis-content-quality |
v4.0 slop-detection methodology: model-family fingerprinting (8 families), substance-and-depth tests, two-axis calibration, compounding archive |
synthesis-fact-checking |
v2.0 fact-checking with 9 new protocols: nested attribution, composite quotes, paraphrase drift, citation laundering, AI-synthetic sources, tool-specific hallucination signatures |
synthesis-article-refresh |
Refresh old blog posts while maintaining temporal integrity |
| Skill | Description |
|---|---|
synthesis-concise-messaging |
High-Five Habit — condense messages to 5 sentences or less |
| Skill | Description |
|---|---|
synthesis-context-lifecycle |
Three-tier context architecture for managing AI working memory |
synthesis-project-management |
Lightweight PM system for human-agent collaboration |
synthesis-daily-rituals |
Day-start and day-end checklists with dependency-ordered rituals |
| Skill | Description |
|---|---|
synthesis-anti-shortcuts |
Deterministic enforcement of anti-shortcut discipline: costume-vocabulary catalog, constraint-first protocol, sub-agent hygiene, case studies |
synthesis-content-framing |
Content framing with topic, sophistication, and engagement gates |
| Skill | Description |
|---|---|
synthesis-thinking-framework |
Five-mode thinking methodology: first principles, systems, complexity, analogical, and design thinking |
synthesis-voice-profiler |
Generate a structured writing voice profile from samples for agent instruction files |
synthesis-tree-of-thought |
Multi-expert collaborative reasoning technique |
synthesis-llm-setup |
Configure Claude Projects, ChatGPT GPTs, and Gemini Gems |
synthesis-creative-writer |
Creative writer persona template |
synthesis-technical-advisor |
Technical advisor persona template |
| Skill | Description |
|---|---|
synthesis-git-hooks |
YAML-driven pre-commit policy: auto-classifies each repo by push remotes (personal vs strict), enforces tiered patterns for credentials and exposure-sensitive content |
synthesis-inbox-cleanup |
Manifest-driven email cleanup across iCloud / generic IMAP (Python), Microsoft 365 / outlook.com (Mail.app AppleScript), and Gmail (workspace-mcp API + native server-side filters). Public engine + private rules. Ships with prompt-injection defenses and adversarial test fixtures for any LLM-augmented path. macOS. |
synthesis-mac-sync |
Multi-Mac config sync via iCloud with git repo sync and machine inventory |
synthesis-meeting-transcripts |
Fetch AI-generated meeting notes and transcripts into local working files |
synthesis-repo-guard |
Session-end enforcement: detect uncommitted changes and unpushed commits across a workspace |
synthesis-slack-sync |
Slack channel sync protocol: read channels, threads, DMs to local transcripts |
synthesis-skills-manager |
Agent-native skill installer: drift detection, synthesis merge, provenance tracking |
| Skill | Description |
|---|---|
synthesis-clean-text |
Produce text without AI watermarking patterns |
synthesis-response-merger |
Combine multiple LLM responses into one unified document |
Skills use progressive disclosure:
- Tier 1 (always loaded): name + description (~50 tokens) — matches your requests
- Tier 2 (on activation): SKILL.md body — the actual instructions
- Tier 3 (on demand): reference files for detailed material
When you ask your AI assistant to do something that matches a skill's description, it loads automatically. Skills that involve writing include defaults that work standalone. If you have personal preferences in agent instruction files such as CLAUDE.md or AGENTS.md, those override the defaults.
Many of these skills are practical artifacts of synthesis engineering, including synthesis coding, synthesis writing, and synthesis project management.
- CC0 1.0 — methodology and content skills (no attribution required)
- Apache 2.0 — skills with executable scripts
Read the launch article: Synthesis Skills: Install Methodology Into Your AI Workflow
- Synthesis coding — AI-assisted software development
- Synthesis engineering — broader human-AI collaboration methodology
- Agent Skills standard — the open format these skills use
See CONTRIBUTING.md.
Rajiv Pant — technology executive, AI practitioner, and creator of synthesis coding.