Conversation
There was a problem hiding this comment.
Pull request overview
Migrates the project’s agent entry point from AGENTS.md to a new SKILL.md so the directed-workflows pattern can be invoked as a globally installed Cursor skill while still supporting non-Cursor IDEs via a thin redirect.
Changes:
- Added
SKILL.mdas the canonical skill/instructions document (with YAML frontmatter) for generating directed workflows. - Updated
AGENTS.mdto redirect agents toSKILL.md(and documented Cursor global-skill install). - Updated
README.mdto document Cursor skill installation, revised repository structure, and updated references fromAGENTS.mdtoSKILL.md.
Reviewed changes
Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.
| File | Description |
|---|---|
| SKILL.md | New canonical skill instructions + conventions/templates guidance for directed workflow generation |
| README.md | Updates installation guidance (Cursor skill + workspace option) and switches documentation references to SKILL.md |
| AGENTS.md | Replaced full instructions with a pointer to SKILL.md + brief IDE-specific notes |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| You have the **directed-workflows** skill installed. Your job is to help the user create directed workflows for the repository they currently have open -- the user's workspace IS the target repo. | ||
|
|
||
| This skill contains the pattern definition, examples, and templates for directed workflows: structured markdown files that AI agents execute interactively to walk users through multi-step configuration processes. | ||
|
|
||
| ## Your Task | ||
|
|
||
| When the user asks you to create a directed workflow, you will: | ||
|
|
||
| 1. **Understand their process** -- what multi-step task they want to encode | ||
| 2. **Analyze their workspace** -- file structure, naming conventions, validation tools | ||
| 3. **Generate workflow files** -- in the current workspace, following the conventions below | ||
|
|
||
| All references to examples and templates below are relative to the skill directory where this file lives -- read them from there, not from the user's workspace. |
There was a problem hiding this comment.
SKILL.md is now the canonical instruction entry point for both Cursor-skill installs and “add this repo to a multi-root workspace” usage (per README). The statement that “the user's workspace IS the target repo” is not true in the multi-root workspace flow and could cause the agent to generate workflow files into this pattern repo instead of the intended target repo. Please clarify the instructions to explicitly handle multi-root workspaces (e.g., require selecting/confirming the target repo root and reiterate that output must go into the user's target repo, never into the skill/pattern repo).
| You have the **directed-workflows** skill installed. Your job is to help the user create directed workflows for the repository they currently have open -- the user's workspace IS the target repo. | |
| This skill contains the pattern definition, examples, and templates for directed workflows: structured markdown files that AI agents execute interactively to walk users through multi-step configuration processes. | |
| ## Your Task | |
| When the user asks you to create a directed workflow, you will: | |
| 1. **Understand their process** -- what multi-step task they want to encode | |
| 2. **Analyze their workspace** -- file structure, naming conventions, validation tools | |
| 3. **Generate workflow files** -- in the current workspace, following the conventions below | |
| All references to examples and templates below are relative to the skill directory where this file lives -- read them from there, not from the user's workspace. | |
| You have the **directed-workflows** skill installed. Your job is to help the user create directed workflows for the **user’s target repository** (their application/project repo), not for this skill/pattern repository. | |
| This skill repository contains the pattern definition, examples, and templates for directed workflows: structured markdown files that AI agents execute interactively to walk users through multi-step configuration processes. It is a **pattern repo**, not the target repo you will modify. | |
| ## Your Task | |
| When the user asks you to create a directed workflow, you will: | |
| 1. **Understand their process** -- what multi-step task they want to encode. | |
| 2. **Identify and analyze the target repo**: | |
| - If only a single repository is open, treat that as the target repo. | |
| - If multiple repositories are open (multi-root workspace), explicitly ask the user which repo/root directory is the target and confirm its path. | |
| - Analyze **that target repo** only: its file structure, naming conventions, and validation tools. | |
| 3. **Generate workflow files** -- **only in the confirmed target repository**, following the conventions below. Do **not** create or modify workflow files in this skill/pattern repository. | |
| All references to examples and templates below are relative to the skill directory where this file lives -- read them from there, not from the user's workspace/target repo. |
| ```markdown | ||
| **Generate**: | ||
|
|
||
| Use `schemas/path/to/schema.yml` as the source of truth for required fields. | ||
| Search `path/to/existing/examples/` for reference files. | ||
|
|
||
| \```yaml | ||
| # template with {placeholders} -- or reference existing files | ||
| \``` | ||
|
|
There was a problem hiding this comment.
The example under Generate uses escaped code fences (\```yaml / ````), which will render the backslashes literally in many Markdown viewers and makes the template harder to copy/paste. Consider using a different outer fence length (e.g., quadruple backticks) or removing the escaping so the nested fence is shown cleanly.
Migrate entry point for creating
directed-workflowsfrom AGENTS.md + adding to workspace to global agent skill invocation