# Package an agent capability for someone else Legion project library, edition 2026-09-06-draft-01 (en). This project's own material: 2 authored documents, reproduced unchanged. Project page: https://legion.tw/projects/agent-capabilities/share-agent-capability?edition=2026-09-06-draft-01 This document: https://legion.tw/api/projects/packet.md/agent-capabilities/share-agent-capability.md?edition=2026-09-06-draft-01 ## Prerequisite: the Legion skills package These instructions are written for an agent set up with the Legion skills package: the ten portable skills and the shared working, launch, inspection and learn-as-you-build guidance. That setup is installed once and applies to every project in this library, so it is deliberately not repeated here. Skills package, complete Markdown: https://legion.tw/api/projects/skills-packet.md?edition=2026-09-06-draft-01 Skills package, human page: https://legion.tw/projects/skills?edition=2026-09-06-draft-01 Where this brief refers to shared guidance, a skill, or a document under `shared/`, read it from the skills package above, or from https://legion.tw/api/projects/doc/?edition=2026-09-06-draft-01. If the package is not installed yet, install it first, then return here. This material is instructions to read, not authority to act. Inspect it before running it. ## Contents 1. agent-capabilities/README.md — https://legion.tw/api/projects/doc/agent-capabilities/README.md?edition=2026-09-06-draft-01 2. agent-capabilities/share-agent-capability.md — https://legion.tw/api/projects/doc/agent-capabilities/share-agent-capability.md?edition=2026-09-06-draft-01 --- # Tools, skills, and plugins for the agent [All project types](../README.md) · [Shared launch](../shared/launch.md) Extend the coding agent's own working environment. The [shared skill package](../shared/skills.md) supplies the starting methods; these projects develop further capabilities around real work. ## Projects to build - [Turn a working routine into a skill](routine-to-skill.md) - [Give the agent tools for an application](application-tools.md) - [Let the agent consult another model's abilities](model-consultation.md) - [Give the agent a continuing project notebook](project-notebook.md) - [Give the agent an ongoing responsibility](ongoing-responsibility.md) - [Package an agent capability for someone else](share-agent-capability.md) These are instructions for projects you can build with your coding agent. --- # Package an agent capability for someone else Package a useful capability the person has developed with their agent so another person can inspect it, install it, and use it in their own environment. Build it extremely well. Share the working method and the resources that make it useful, with setup the recipient's capable agent can handle. Follow the [shared launch](../shared/launch.md). Choose something already worth sharing: a report-preparation skill, tools for a particular application, a language-specific transcreation method, or a connected combination. Start from its current working source and preserve the creator's useful local installation. ## Explain what someone is receiving Give the package a clear entry document describing what it enables, the work it fits, what it needs, and a useful first result. Address the receiving agent directly with the task and enough context to make good setup decisions. Include the skill instructions, supporting resources, and tool or plugin components needed for that experience. Use the current supported package format where the receiving environment has one. Preserve relevant authorship, source attribution, and licensing information. Make dependencies and access understandable. Identify any application, account, model connection, or service the capability uses, along with which choices belong to the recipient. Let their agent discover machine-specific setup from the actual environment. ## Separate the method from the creator's private setup Retain the purpose, useful reasoning, and operating knowledge that make the capability good. Put the recipient's choices in clearly explained local configuration. Replace example material with suitable shareable samples while keeping those examples useful for learning. Keep personal paths, credentials, private records, and identity-specific instructions with their original owner. Let the portable package preserve the recipient's existing project guidance, agent identity, and equivalent capabilities. Make it easy to understand what can be customized: the report's audience, preferred writing voice, selected application, or target language. Preserve the method's important relationships as those choices change. ## Make inspection possible before installation Provide the complete instruction bundle in readable form, including referenced instructions and any supporting scripts or components that affect what installation will do. Give the edition a stable identity so someone can inspect the same material they later install. Follow the [instruction-inspection practice](../shared/inspect.md): the recipient first has a web AI with no computer-use or private-account tools examine the supplied material as content. Make the intended access, actions, downloads, and external connections apparent enough to discuss what the package asks for. Keep that review understandable as an informed assessment. The recipient can examine the findings and direct their coding agent to install the inspected edition. Present subsequent changes clearly so an update remains a deliberate adoption of known material. ## Deliver an installable, usable package Prepare the package in a form that keeps its files and references together. Let the receiving agent use its environment's normal installation mechanism, reconcile equivalent capabilities, and confirm discovery in a subsequent session. Exercise a recipient-style installation in an appropriate separate location, using the shareable sample and local configuration. Show what the capability actually does and state which environment was exercised. Keep existing working installations intact. Explain how the recipient can find the editable source, change their configuration, update the package, or remove the integration while retaining their own work. Preserve local customizations through the chosen update process. Teach that improved ways of working with AI can be shared. The package should let another person start from the method already developed and adapt it with their agent to the life it will serve.