$ ~/registry/skill/microsoft-adapt-skill

SKILL

adapt-skill

Name: adapt-skill
Author: microsoft

Convert skills from other AI coding assistants into Amplifier-native SKILL.md files.

Stars

★ 9

Source

GitHub

Updated

2026-07-09

// securityCaution

Prompt only, no code
Open source

Summary

This skill is an open-source, prompt-only adaptation workflow with no required secrets, no declared fixed remote endpoints, and no built-in code execution capability. Overall risk is low, with the main caveat being low community adoption, no declared license, and unknown maintenance status.

CredentialsLow risk

The material explicitly states that no keys or environment variables are required. There is no indication that the skill requests API keys, tokens, or other sensitive credentials, nor any credential storage or forwarding logic.

Network egressLow risk

No remote host is declared. Although the README mentions optional use of web search/web_fetch to research source platform conventions, it does not bind to any unknown third-party service or show a default fixed path for sending user data out.

Code executionLow risk

The system flags it as 'prompt-only', indicating it is an instruction/workflow skill rather than executable software. The material shows no install script, local process spawning, shell execution, or request for elevated system permissions.

Data accessLow risk

The described function is mainly to read a source skill file, analyze its structure, and produce an adapted SKILL.md. It does not claim broad scanning, persistent collection, or writing beyond what the task requires. As prompt-only content, any actual data access remains subject to host-platform permissions.

Supply chainCaution

The source is an open-source GitHub repository, which is a clear risk-reducing factor because the code is auditable. However, it has 0 stars, no declared license, and unknown maintenance status, so community validation and ongoing maintenance signals are weak and warrant caution before production use.

Recommendations

Validate its adaptation output in a restricted test environment before using it for production skill publication.
Manually review the generated SKILL.md, especially mapped tool names, frontmatter fields, and any permission-related instructions.
Confirm ongoing maintenance and licensing status, and pin to an audited commit if you decide to use it.

Audited by: gpt-5.4 · 2026-06-16

// install

Copy the install command and let the AI configure it · recommended for beginners

Please install the "adapt-skill" skill from askskill:
1. Download https://raw.githubusercontent.com/microsoft/amplifier-bundle-skills/main/skills/adapt-skill/SKILL.md
2. Save it as ~/.claude/skills/adapt-skill/SKILL.md
3. Reload skills and tell me it's ready

// download

Download SKILL.mdInstall manifest ↗

// examples

Migrate a Claude Code skill

Prompt

Please convert the following skill written for Claude Code into an Amplifier-compliant SKILL.md. Preserve the original intent and point out any platform-specific conventions that need adjustment:

[paste the source skill content]

Expected

A complete Amplifier SKILL.md plus notes on Claude Code-specific parts that were replaced or rewritten.

Convert a Cursor skill template

Prompt

I have a Cursor skill template. Adapt it into an Amplifier-native skill file, fill in any missing sections, and ensure it follows the Agent Skills spec with Amplifier extensions:

[paste the skill template]

Expected

A ready-to-use Amplifier skill document with normalized structure and completed required sections.

Annotated skill adaptation

Prompt

Please convert this SKILL.md from another AI assistant platform into an Amplifier version, and add brief notes for each major change so I can understand why it was modified:

[paste the original]

Expected

An Amplifier version of the SKILL.md with major-change annotations for easier migration and review.

// docs

Adapt Skill

Adapt a skill from another AI coding assistant into a properly structured Amplifier SKILL.md. The output must conform to the Agent Skills specification with Amplifier extensions and be immediately usable via /skill-name.

Inputs

$ARGUMENTS: (Optional) Path to the source skill file or description of what to adapt.

Steps

1. Consult Skills-Assist

Before writing any skill content, load the authoritative skills reference:

load_skill("skills-assist")

Ask skills-assist about:

Current frontmatter field reference and best practices
Fork vs inline decision criteria
Step annotation conventions
Testing patterns for the type of skill being created

skills-assist is the source of truth for Amplifier skill conventions. The examples in this skill are illustrative samples — consult skills-assist for the complete and up-to-date specification.

Success criteria: You have loaded and consulted skills-assist for the latest skill authoring conventions.

2. Explore the Target Bundle

If the user wants to add the skill to a bundle (e.g., amplifier-bundle-skills), delegate to foundation:explorer to understand the bundle's conventions, existing skills, and structure. If saving to a personal or project directory, skip this step.

Success criteria: You understand the target bundle's skill conventions and can produce a skill that fits in.

3. Read and Analyze the Source Skill

Read the source SKILL.md and identify:

The skill's purpose and core workflow
Platform-specific tool names (e.g., Read vs read_file, AskUserQuestion vs natural conversation, Bash(mkdir:*) vs bash)
Platform-specific paths (e.g., .claude/skills/ vs .amplifier/skills/)
Platform-specific frontmatter fields that need mapping to Amplifier equivalents or removal (consult skills-assist for the current Amplifier field reference)
Conventions that don't exist in Amplifier and need alternative approaches
The interview or interaction model (does it rely on tools Amplifier doesn't have?)

Present a summary of what needs to change — grouped into natural clusters, not a wall of text. Let the user confirm the adaptation plan before proceeding.

Success criteria: You have a clear mapping of source conventions to Amplifier equivalents, and the user has confirmed the approach.

4. Research Source Platform Conventions (if needed)

If the source skill uses unfamiliar fields or conventions, research the source platform's official documentation using web search and web_fetch. Compare the source platform's spec with Amplifier's to identify:

Fields that map directly (e.g., name, description, allowed-tools)
Fields that map with translation (e.g., CC's model -> Amplifier's model_role)
Fields with no Amplifier equivalent (document the gap, decide whether to drop or approximate)

Skip this step if the source conventions are already well understood.

Success criteria: You understand the source platform's skill format well enough to make accurate adaptation decisions.

5. Design the Adapted Skill

Walk through these decisions with the user, grouped into natural clusters (not one question per turn, not everything at once):

Identity and routing:

Suggested name and description (with trigger phrases for the visibility hook — the description is the ONLY signal the model sees before deciding to load)
model_role based on the skill's primary cognitive task (consult skills-assist for the full set of available roles)

Execution model:

Inline vs forked (context: fork). Critical: forked skills CANNOT do multi-turn conversation with the user — the sub-agent runs once and returns. If the skill needs to interview the user, it must be inline.
Model-invocable vs user-only (disable-model-invocation)
allowed-tools — minimum set needed, translated to Amplifier tool names (consult skills-assist for the current tool name reference)

Save location:

…

View full docs ↗