Ask a real question.
Work normally. Auto skills load when your agent recognizes the situation. Command skills wait for you to invoke them.
OPEN SOURCE. OPINIONATED. YOURS.
Better instincts.
For your AI agent.
Give your agent a better way to think.
Practical skills for designing systems, untangling code, and making yourself understood.
Made for Claude Code·Codex CLI·opencode
30↗skills to think with
6disciplines, one collection
3supported agent runtimes
Written once.
Ready for your workflow.
LESS PROMPTING. MORE DOING.
Small files. A different approach.
Bring in the right thinking.
A skill gives your agent a method, a vocabulary, and useful guardrails. References load only when needed.
Make it your own.
Read the Markdown. Keep what works. Change what doesn't. This is a personal collection, built to be inspectable.
THE COLLECTION
Good work starts
with the right skill.
From the shape of your system
to the clarity of your next sentence.
30 skills
Router
ai-ml
Routes work on trained-model systems to the right sub-skill: LLM apps (prompting, RAG/retrieval, agents and tool use, evals and LLM judges, guardrails, fine-tuning) or classical ML projects (framing, baselines, model choice, missing data, deployment, retraining).
Router
architecture
Routes software-architecture work to the right sub-skill: choosing topology or code organization (monolith vs microservices, layered/hexagonal/clean), service boundaries and inter-service communication, DDD modeling, component decomposition, C4 diagrams, coupling/cohesion analysis, and CI architecture rules.
Via architecture
architecture-pattern-advisor
Guides choosing a repository's architecture — topology (monolith, modular monolith, microservices, serverless) or code organization (layered, by-domain, hexagonal, clean/onion).
Auto
blue-team
Develops a raw early idea (tech decision, project pitch, LinkedIn thesis, personal choice) into a sharper version that holds up before its audience.
Via architecture
c4-modeling
Drafts a C4 model of a software system with the user and renders it as Mermaid at Context, Container, Component, Landscape, Dynamic, and Deployment level (not Code).
Auto
communication-analysis
Analyzes or rewrites a message, feedback, or conversation with Schulz von Thun's four-sides model and Rogers-style active listening.
Via architecture
coupling-cohesion
Assesses existing code for module cohesion and LCOM, codebase-wide coupling metrics and zones, or the balance of one dependency (Khononov).
Via architecture
ddd
Applies Domain-Driven Design: subdomain classification, context mapping, legacy integration (ACL), tactical pattern choice, and conventions for aggregates, value objects, events, and event sourcing.
Auto
documentation
Guides writing or revising technical docs for a named reader — README, tutorial, how-to, API reference, docstrings, runbook, package docs, or vignettes.
Via architecture
fitness-functions
Designs and implements architecture fitness functions — automated CI checks that fail the build when a characteristic such as modularity, layering, coupling, security, or resilience erodes.
Auto
html-artifacts
Produces a self-contained HTML file for deliverables the reader will keep, share, or manipulate; not for ordinary answers.
Auto
hypertrophy-training
Guides experienced trainees in hypertrophy programming, auto-regulation, and plateau diagnosis.
Command
improve-codebase
Scan a codebase for deepening opportunities, present them as a visual HTML report with before/after diagrams, then interview the user round by round about the candidate they pick.
Auto
ki-stil
Checks a German draft for wording that reads generic or AI-typical, with counted, located signals and rewrites.
Via ai-ml
llm-application-engineering
Engineers LLM applications: fixing bad output (prompting, RAG, finetuning), evals and LLM judges, agents and tools, guardrails, and production monitoring.
Via architecture
logical-component-design
Decomposes a NEW system or feature into named logical components; not for measuring an existing decomposition.
Via architecture
microservices-design
Guides microservice interaction design and review — boundaries, coupling, communication style, contract versioning, sagas, and resiliency patterns from Newman's *Building Microservices*.
Via ai-ml
ml-project-lifecycle
Guides a classical ML project — framing, baselines, model choice, missing data, deployment, retraining.
Auto
natural-planning
Plans one multi-step project with GTD's Natural Planning Model (purpose, outcome, brainstorm, organize, next actions).
Command
oss-scouting
Scout one third-party open-source repository for open issues that suit a small, clean contribution, and write repro, root-cause analysis, fix diff, test, and a submit checklist to a local folder only.
Command
poc-spec-loop
Takes a greenfield R, Python, or bash PoC to production readiness in two gated phases — interactive spec, then a per-task TDD loop with fresh context up to a pull request.
Auto
problem-first-explanation
Structures explanations of tools, methods, concepts, rules, and proposals by leading with the concrete problem.
Command
pypet-snippets
Curate shell-command snippets with the pypet CLI — create, find, edit, run and alias them on request, including proposing one from commands that recur in shell history.
Auto
red-team
Red-teams a finished or near-finished artefact or decision with a materiality gate and a verdict.
Auto
refactoring
Finds where refactoring is worth starting in an unfamiliar codebase by ranking files by git churn (hotspots) and explaining the ranking.
Command
release-pr
Turn the current branch into a release pull request for an R or Python package — confirmed version bump, finalized NEWS.md or CHANGELOG.md, checks run, PR body written.
Command
repo-status
Generate a standup / status update from recent development activity — yesterday/today/blockers, turning rough notes or connected-tool activity into a shareable update.
Auto
stakeholder-update
Drafts a weekly update, executive summary, escalation, or launch announcement for readers outside the working group.
Command
tldr
Compressing something too long to read into the few facts needed to decide or act, read instead of the source — the last message and the work behind it, or a named file, PR, document, or thread.
Command
was
Re-pitching a short message that was read but not understood, in plain German — the missing context added first instead of shortening, short active sentences, the repo's glossary terms kept verbatim.
No skills found.
Try a broader search or another discipline.
YOUR AGENT. WITH BETTER HABITS.
Make room for
better thinking.
Pick your agent. Add the skills.
Get back to the work that matters.
Claude Code / marketplace
/plugin marketplace add fabiandistler/agents
/plugin install architecture@fabiandistler-agentsStart with architecture. Install other category plugins when you need them. For Claude Desktop, claude.ai, and Cowork, use the plugin guide.
Codex CLI / local skills
git clone https://github.com/fabiandistler/agents.git ~/src/agents
cd ~/src/agents
./install.sh --target=codexRequires Bash, Git, and Python 3. The skills stay linked to your local clone. See the guide for updates and optional guard hooks.
opencode / local skills
git clone https://github.com/fabiandistler/agents.git ~/src/agents
cd ~/src/agents
./install.sh --target=opencodeRequires Bash, Git, and Python 3. The skills stay linked to your local clone. See the guide for updates and optional guard hooks.