Skip to main content

Socratic: Set up accepted agent rules for a

Interview the user about a repository and connected agent tools, then write only the harness rules and Skills they accept.

Source facts

Repository
beelarr/soc
Last source activity
October 6, 2026 at 23:08
Detected SKILL.md language
English
Stars
0
Forks
0

Uses

Use it after installation to set up a repository’s agent instructions or to bring those instructions up to date.

Prerequisites

Point the Skill at the repository to configure. It asks which agent tools to connect and reads the repository’s manifests and code before proposing rules.

How to use

Provide the target repository and answer the Skill’s questions. It follows its setup, research, interview and verification steps before writing accepted files.

Limitations

The source forbids deleting existing source files or Git metadata and requires the user’s approval before installing a Skill, server or package.

Install options

The review-first prompt is selected by default. You can switch to a direct command or download a local copy.

Review the source files

Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.

File Explorer
5 files

Showing SKILL.md

SKILL.md
Source instructions · Read-only preview
name
socratic
description
Set up a repository after this skill is installed. Asks which agent tools to connect, reads the manifests and the code, and writes rules and skills the person accepts. Use when the person installs this skill, wants harness instructions, or wants the harness instructions brought up to date.
# Socratic You are already the agent. The person installed this skill. This directory is not the repository to wrap. Do not inspect this skill's own folder for the project's manifests, and do not clone this repository. If they just installed you, set the repository up before you do anything else. ## Workspace Ask which repository to wrap if the person has not already pointed at one. Use its absolute path for the rest of the session. Call that path `WORKSPACE`. ## The steps Four steps ship in this folder. They are not separate skills. Read each file and follow it, in this order: 1. `setup.md` 2. `research.md` 3. `interview.md` 4. `verify.md` If one is missing, stop. Do not invent a replacement. Do not run a separate program. There is no CLI. ## Rules for every step - Do not delete, move, or overwrite a file this skill did not write. You may edit a rule, skill, or MCP entry this skill wrote. You may not delete source files, manifests, `.git`, or a tool directory that was already there. - Do not create, delete, or modify `.git`. - Do not invent a practice the person did not state. An answer that does not change how an agent should work produces no file. - Do not install a skill, an MCP server, or a package until the person says yes in this chat. Do not treat silence as a yes. - Do not read `.env`, `.env.*`, `*.pem`, or `*.key`. - Do not upload the repository. A documentation fetch uses a URL you show the person first.
View on GitHub