create-agent
Scaffolds platform-native Cursor Markdown or Codex TOML agent definitions through six layers: Role, Inputs, Process, Output contract, Constraints, and…
Skill by dasobral
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/dasobral-create-agent
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: documents, agent-building
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/create-agent
What's inside
- Skill name
create-agent- Description
- Scaffolds platform-native Cursor Markdown or Codex TOML agent definitions through six layers: Role, Inputs, Process, Output contract, Constraints, and Integration hooks. Part of Agent Platform plugin. Use when creating agents or formalizing multi-agent components.
- License
- MIT
- Instructions
- 84 lines in SKILL.md · view on GitHub
Files (5)
- SKILL.md
- summarizer.md · examples/
- agent-conventions.md · references/
- agent-definition.md · templates/
- agent-definition.toml · templates/
SKILL.md (first lines)
# Create Agent Scaffold a complete native agent definition for Cursor or Codex. ## Step 1 — Resolve Platform, Name, and Path Use the user's explicit target when provided. Otherwise infer the platform only from unambiguous repository configuration. For an unknown platform, ask which target to use. Normalize the name to `lowercase-kebab-case`, then select the native path: | Platform | Definition | Default path | |----------|------------|--------------| | Cursor | Markdown | `./agents/<name>.md` or the plugin's `agents/` | | Codex | TOML | `.codex/agents/<name>.toml` | ## Step 2 — Load Conventions Read `references/agent-conventions.md`, then load the native template: - Cursor: `templates/agent-definition.md` - Codex: `templates/agent-definition.toml` ## Step 3 — Gather Six Layers | Layer | Capture | |-------|---------| | Role | Identity, purpose, tone, scope | | Inputs | name, type, required, description per parameter | | Process | Numbered imperative steps (3–15) | | Output | One value from the exact output vocabulary | | Constraints | Read-only, scope limits, tool restrictions | | Integration | Parent orchestrator, artifact routing, dependency handoffs | The exact output vocabulary is: `json_object`, `markdown_document`, `file_write`, `inline_text`, or `tool_call_sequence`. ## Step 4 — Render the Native Definition For Cursor, use Markdown with frontmatter: ```yaml --- name: agent-name description: One-line trigger description for agent discovery. model: fast # optional: fast | inherit | specific model slug --- ``` For Codex, write valid TOML with all required keys: ```toml name = "agent-name" description = "One-line discovery description." developer_instructions = """ State the role, inputs, process, output contract, constraints, and handoffs. """ ``` Do not hard-code a Codex model or sandbox policy. Put the complete behavioral contract in `developer_instructions`. ## Step 5 — Write File Write the complete definition to the resolved native path. Include an example from `examples/` only for Cursor when helpful. ## Step 6 — Validate - [ ] File parses as Markdown frontmatter or TOML for the selected platform - [ ] Definition has non-empty `name` and `description` - [ ] Codex definition has non-empty `developer_instructions` - [ ] At least one input parameter defined - [ ] Output contract is machine-parseable - [ ] Integration contract defines parent and artifact handoffs ## Rules - Match the `orchestrate` skill's handoff contract without assuming a specific delegation API - Preserve platform-native syntax and paths - See `references/agent-conventions.md` for naming and structure
From dasobral/skills, MIT.
More from dasobral/skills
- 0.529score
analyze-codebase
dasobral
Analyzes a codebase across 16 dimensions — naming, formatting, architecture, async patterns, error handling, testing, anti-patterns — and writes…
SkillBFreshness: BClaude Code +9 - 0.529score
agent-ste
dasobral
Use when rewriting user or planner instructions before planning or execution; when prompts contain vague verbs (optimize, improve, better, appropriate…
SkillBFreshness: BClaude Code +9 - 0.529score
audit-cognitive-debt
dasobral
Audits cognitive load across seven debt categories: dead code, over-engineering, duplication, abstraction quality, control flow, naming, and test debt.
SkillBFreshness: BClaude Code +9 - 0.529score
cpp-engineer
dasobral
Production-grade C++ for industrial/defense QKD ground-segment systems.
SkillBFreshness: BClaude Code +9 - 0.529score
cpp-review
dasobral
Unified C++ review for QKD/defense systems with three lenses: --realtime (concurrency, latency, RAII), --security (crypto hygiene, ETSI, threat model), --full…
SkillBFreshness: BClaude Code +9 - 0.529score
orchestrate
dasobral
Top-level orchestrator that deploys specialized agents from an agents/ directory to complete complex tasks.
SkillBFreshness: BClaude Code +9 - 0.529score
local-inference
dasobral
Local LLM inference setup and optimization for agentic workflows.
SkillBFreshness: BClaude Code +9 - 0.526score
build-attack-scenario
dasobral
Build a portable, hash-pinned agent attack scenario with explicit trusted goals, untrusted channels, prohibited effects, and benign criteria.
SkillBFreshness: BClaude Code +9 - 0.526score
assess-repository-trust
dasobral
Inventory repository control-plane inputs and produce a private, hash-based trust record before an agent executes repository-supplied automation.
SkillBFreshness: BClaude Code +9 - 0.525score
career-documents
dasobral
Write and edit professional career documents: CVs, cover letters, research statements, fellowship proposals, and resignation letters.
SkillBFreshness: BClaude Code +9 - 0.523score
challenge-sciml-model
dasobral
Challenge SciML evidence for leakage, physical constraints, extrapolation, uncertainty, baselines, and cost.
SkillBFreshness: BClaude Code +9 - 0.523score
audit-scientific-claim
dasobral
Audit claims against hashed evidence while separating verification, validation, and uncertainty quantification.
SkillBFreshness: BClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Aug 15, 2026 (2 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/dasobral/skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from dasobral
- 0.529score
local-inference
dasobral
Local LLM inference setup and optimization for agentic workflows.
SkillBFreshness: BClaude Code +9 - 0.529score
cpp-review
dasobral
Unified C++ review for QKD/defense systems with three lenses: --realtime (concurrency, latency, RAII), --security (crypto hygiene, ETSI, threat model), --full…
SkillBFreshness: BClaude Code +9 - 0.529score
audit-cognitive-debt
dasobral
Audits cognitive load across seven debt categories: dead code, over-engineering, duplication, abstraction quality, control flow, naming, and test debt.
SkillBFreshness: BClaude Code +9 - 0.529score
orchestrate
dasobral
Top-level orchestrator that deploys specialized agents from an agents/ directory to complete complex tasks.
SkillBFreshness: BClaude Code +9 - 0.529score
cpp-engineer
dasobral
Production-grade C++ for industrial/defense QKD ground-segment systems.
SkillBFreshness: BClaude Code +9 - 0.529score
agent-ste
dasobral
Use when rewriting user or planner instructions before planning or execution; when prompts contain vague verbs (optimize, improve, better, appropriate…
SkillBFreshness: BClaude Code +9
Cite this page
create-agent skill. Agent Mart. https://agentmart.dev/skills/dasobral-create-agent (last updated 2026-10-24)