create-agent
Skill0 starsChanged 7 months agoMITView on GitHub
Use when the user says "/create-agent [name]", "neuen Agent erstellen", "create agent", or asks to scaffold a new Claude Code subagent. Creates an agent definition file in `.claude/agents/<name>.md` following current Anthropic best practices (Opus 4.7 / 2026), including frontmatter schema and minimum-permission tools.
- Deletes or force-pushes
38 other projects use a version of this.
Adopted by
Most recently taken up by
- lalulali/vespyrtheir own version, first seen Nov 6, 2026
- cognitedata/builder-skillstheir own version, first seen Nov 4, 2026
- harness/harness-skillstheir own version, first seen Nov 1, 2026
- fxmartin/claude-code-configtheir own version, first seen Oct 31, 2026
- agno-agi/agentos-dockertheir own version, first seen Oct 30, 2026
- agno-agi/agentos-flytheir own version, first seen Oct 30, 2026
Used with
Projects with it also use deploy-platform, setup-platform and 8 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
agno-agiSkilldeploy-platform
Deploy this AgentOS using its own provider instructions, configure JWT verification, and verify the public service. Use for first deployment, redeployment, or production configuration changes.
Also in 5 projects that use this
agno-agiSkillsetup-platform
Set up this AgentOS from a fresh clone — confirm Docker, configure .env, boot the containers, prove the MCP endpoint live, connect the AgentOS UI, then build the user's first agent. Use when the user asks to set up the platform, get started, or bring this repo up on a new machine.
Also in 5 projects that use this
ElasticSkillcreate-integration
Use when creating a new Elastic integration package, scaffolding data streams, answering package layout or structure questions, or running the end-to-end integration build workflow. Covers package topology, scaffold commands, post-scaffold edits, and full orchestration of CEL/pipeline/test subagents.
Also in 3 projects that use this
MicrosoftSkillcreate-hook
Create a hook (.json) to enforce policy or automate agent lifecycle events.
Also in 5 projects that use this
AnasX7Skillimprove
Survey any codebase as a senior advisor and produce prioritized, self-contained implementation plans for OTHER models/agents to execute. Strictly read-only on source code — never implements, fixes, or refactors anything itself. Use when asked to audit a codebase, find improvement opportunities (bugs, security, performance, test coverage, tech debt, migrations, DX), suggest features or where to take the project next (roadmap, product direction), or generate handoff plans for another agent to implement.
Also in 5 projects that use this
SentrySkillcreate-pr
Create pull requests following Sentry conventions. Use when opening PRs, writing PR descriptions, or preparing changes for review. Follows Sentry's code review guidelines.
Also in 3 projects that use this
- +3 more from agno-agi
Other versions38 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 38
- microsoft/vscode★ 195ksame name
- gnekt/My-Brain-Is-Full-Crew★ 3.9ksame name
- EvolutionAPI/evo-nexus★ 545same name
- evolution-foundation/evo-nexus★ 535same name
- openma-ai/open-managed-agents★ 348same name
- luanmorenommaciel/agentspec★ 256same name
- harness/harness-skills★ 127same name
- microsoft/copilot-studio-plugin★ 82same name
- Azure-Samples/art-voice-agent-accelerator★ 74same name
- TOKTOKHAN-DEV/agent-company★ 74same name
28 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository10 other files it gives its agents.
- accessibility-a11ySkill
- bulletproof-containerSkill
- claude-features-updateSkill
- create-skillSkill
- dokumentenaustauschSkill
- find-skillsSkill
- owasp-securitySkill
- projekt-audit-kundenplanSkill
- session-endSkill
- setup-claude-envSkill
The file
Files
Agent erstellen
Erstellt eine neue Claude Code Subagent-Definition mit aktueller
Anthropic-Konvention. Subagents liegen als einzelne Markdown-Datei in
.claude/agents/<name>.md (global oder projekt-lokal) und koennen vom
Hauptagenten per Agent({ subagent_type: "<name>" ... }) delegiert
werden.
Unterschied zu Skills
- Skill: wird per Trigger-Phrase oder
/commandaktiviert, arbeitet im Haupt-Conversation-Scope, liefert Anleitung an Claude. - Agent/Subagent: separater ausfuehrender Kontext, wird explizit per
Agent-Tool aufgerufen, oft mit eigenem System-Prompt, eigenen Tools und eigenem Modell. Gut fuer Parallelarbeit und Kontext-Isolation.
Argumente
- name — kebab-case, eindeutig (z.B.
ssl-manager) - purpose — 1 Satz, was der Agent autonom erledigen soll
- scope — global (
~/.claude/agents/) oder projekt-lokal (<project>/.claude/agents/) - allowed-tools — Minimum-Set, nur was der Agent wirklich braucht
- model — optional; wenn der Agent ein anderes Modell als die Session verwenden soll (z.B. Haiku fuer schnelle Recherche)
Verzeichnisstruktur
<scope>/agents/<name>.mdAgent-Definitionen sind einzelne .md-Dateien, keine Verzeichnisse. Wenn Hilfs-Skripte oder Referenzen noetig sind, liegen sie daneben oder werden im Agent-Body verlinkt.
Frontmatter-Schema
---
name: <kebab-case> # optional — default ist Dateiname
description: <Wann Agent einsetzen, 1-2 Saetze>
model: claude-haiku-4-5-20251001 # optional — default ist Session-Modell
tools: Read, Grep, Glob, Bash(git log:*) # optional — Minimum-Tool-Set
color: cyan # optional — UI-Akzentfarbe
---Nur setzen, was gebraucht wird. description ist wichtig fuer
Auto-Delegation: der Hauptagent matcht die Aufgabenbeschreibung gegen
description, um zu entscheiden, welcher Subagent geeignet ist.
description-Heuristik
Gut (klar abgegrenzter Zweck + Positiv-/Negativbeispiele):
description: Use this agent to investigate build failures. Reads CI logs, git diff, and recent commits; produces a short root-cause report. Do NOT use for runtime bugs in production — use the log-debug agent for that.Weniger gut:
description: Helps with stuff.Agent-Body-Template
Der Body ist der System-Prompt des Subagents. Er laeuft in isoliertem Kontext, sieht NICHT die aktuelle Conversation. Alles, was er wissen muss, muss im Body stehen.
---
name: <name>
description: Use this agent to <konkreter Zweck>. <Grenzen/Negativfall>.
model: <optional>
tools: Read, Grep, Bash(<subcmd>:*)
---
# <Agent-Titel>
Du bist ein spezialisierter Subagent fuer <Aufgabe>.
## Dein Auftrag
<1-2 Saetze: was du tust, was du nicht tust.>
## Kontext, den du selbst holen musst
- Repo-Layout: siehe `CLAUDE.md` im Arbeitsverzeichnis
- Konventionen: <welche Dateien sind fuehrend>
- Abhaengigkeiten: <z.B. API-Endpoints, Service-URLs, Pfade>
## Arbeitsweise
### 1. <Phase 1 — Analyse/Sammlung>
- <Konkrete Befehle>
- <Was du lesen sollst>
### 2. <Phase 2 — Aktion>
- <Was du schreibst/aenderst>
- <Wie du Rueckfragen vermeidest, wo moeglich>
### 3. <Phase 3 — Abschluss>
- <Strukturierter Report zurueck an den Hauptagenten>
- <Keine Smalltalk, keine Wiederholung der Aufgabe>
## Report-Format
Antworte dem Hauptagenten in dieser Struktur:
Findings:
- ...
Actions:
- ...
Open questions / needs approval:
- ...
## Destruktive Aktionen
- Kein automatisches Git-Push.
- Kein `rm -rf`, kein `docker rm -f` ohne expliziten Auftrag im Prompt.
- Bei Unsicherheit: melden, nicht handeln.
## Grenzen
- <Wofuer der Agent nicht zustaendig ist>
- <Wann der Hauptagent es selbst machen sollte>Workflow
- Agent-Name validieren (kebab-case, kein Duplikat):
ls ~/.claude/agents/ <project>/.claude/agents/ 2>/dev/null - Scope mit User klaeren: global vs. projekt-lokal.
- Datei anlegen:
mkdir -p <scope>/agents/ - Agent-Definition gemaess Template schreiben.
- Dem User die fertige Datei zeigen.
- Fragen, ob committed werden soll (kein automatisches
git add/git commit). - Optional: projekt-spezifische
CLAUDE.mdum einen Agent-Eintrag ergaenzen.
Regeln fuer gute Agents
- description klar abgrenzend — enthaelt "Use this agent for X" und idealerweise "Do NOT use for Y".
- tools minimal — Subagents mit zu breitem Tool-Set tendieren zu Overreach.
- Body = System-Prompt — schreiben, als briefe man einen neuen Kollegen, der die aktuelle Conversation nicht kennt.
- Kein shared State zur Haupt-Session annehmen.
- Report-Format vorgeben — der Hauptagent bekommt nur die finale Antwort zurueck, also soll sie strukturiert und kurz sein.
- Destruktive Aktionen nur, wenn der Prompt sie explizit freigibt. Im Zweifel stoppen und melden.
- Model nur setzen, wenn eine Abweichung vom Session-Modell sinnvoll ist (z.B. Haiku fuer schnelle Parallel-Recherchen, Sonnet fuer Code-Reviews).
- Keine Emojis in Agent-Dateien.
Offizielle Referenz
Anthropic Subagents-Dokumentation: https://code.claude.com/docs/en/subagents
Frontmatter-Felder laut Doku: name, description, tools, model,
color.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

