Agensi Skill Authoring
Write a portable SKILL.md in five phases, from purpose through Agensi packaging.
Free
Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+21 moreClaude CodeCursorCodex CLIMuseDots
Manus
Grok
Grok Bot+18 agentsAgensi Skill Authoring
Example session with this skill installed
I want a new SKILL.md for [one job]. Apply Agensi Skill Authoring. Start with Phase 1. Package it for Agensi. Add agents/grok.yaml only if I will load the skill in Grok Build.
- Read your context and instructions
- Compiled the agensi authoring
Excerpt of the output contract (a real skeleton, not a heading summary). Full copy: references/worked-example.md.
---
name: pr-summary-writer
description: "Draft a reviewer-facing pull request summary from the diff. Use when the user says \"summarize this PR\", \"draft the PR description\", or \"write the pull request summary\". Do not use for a line-by-line code review, or for turning a repeated prompt into a skill (prompt-to-skill-converter-with-grok)."
version: "0.1.0"
author: Author Name
license: MIT
tags:
- ai-agents
- documentation
- developer-tools
- productivity
- git
---
# PR Summary Writer
Write the pull request summary a reviewer reads before the diff.
## Procedure
1. Read the diff and the existing PR text in one batch.
2. Draft a title and a body. Stop only if the repo or the base branch is still unclear.
Trigger test delivered beside that file: "Summarize this PR" loads; "Write up what this pull request changes for the reviewer" loads; "Draft the PR description, and keep the title under 70 characters" loads; "Review this PR for bugs line by line" does not; "Turn the PR-summary prompt I keep pasting into a skill" does not.
Connects securely to your tools. The creator never sees your data.
About this skill
Agensi Skill Authoring turns a blank design into a portable SKILL.md: one job, a gated procedure, a real output contract, and packaging.
Phase 1: Purpose and boundaries. Capture the job, audience, trigger phrases, exclusions, and what done looks like. Stop only when scope or success criteria are unclear. Otherwise state assumptions and continue.
Phase 2: Design the procedure. Write numbered steps with inputs and the evidence that ends each step. Keep each fact in one place.
Phase 3: Define the output. Require a SKILL.md skeleton with a kebab-case name and a trigger description, not a heading summary.
Phase 4: Guardrails. Record only failures the procedure does not already prevent, including broken YAML and equal-length fences.
Phase 5: Package and test. Write frontmatter, permissions, and a Quick Start. Add agents/grok.yaml only for Grok Build, with the same version as SKILL.md. Run three prompts that should load and two that should not. Re-read the files and parse the frontmatter before calling the draft done.
Use it for a new skill or a team authoring standard. Repeated prompts, skill audits, and pre-upload checks are out of scope.
- Covers skill structure. Domain work stays in the skill being written.
- Needs a clear job and success criteria, and stops when those are still unknown.
- Repeated prompts, existing-skill audits, and pre-upload checks are out of scope.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
Trust & safety
Security scan
Security scan passed 12 days ago
- Free with an Agensi account
Permissions
File Scopes
- terminal_shell: "Sometimes – inspect example skills or confirm files exist." - read_files: "Yes – existing skills and drafts." - write_files: "Yes – SKILL.md, optional agents/*.yaml and references/." Token-constrained sessions: see references/token-economy-patterns.md.