Skip to content

docs: clarify how the repo's resource types differ in README - #2517

Open
horner wants to merge 2 commits into
github:mainfrom
horner:docs/readme-when-it-loads-table
Open

docs: clarify how the repo's resource types differ in README#2517
horner wants to merge 2 commits into
github:mainfrom
horner:docs/readme-when-it-loads-table

Conversation

@horner

@horner horner commented Aug 2, 2026

Copy link
Copy Markdown

Why

The What's in this repo table is the first thing a newcomer reads, but it described all five listed resource types as interchangeable-sounding bits of configuration. Nothing in it answered the question people actually arrive with: which of these do I want, and why?

Two concrete gaps:

  1. No distinguishing axis. "Custom agents", "coding standards", "task procedures" are all just descriptions of content. They don't explain that these things behave completely differently at runtime.
  2. The tagline promises six resource types — agents, instructions, skills, hooks, workflows, and plugins — but the table listed four of them, plus the cookbook, which the tagline never mentions. Hooks and workflows were introduced and then never shown.

What changed

Added a "When it loads" column. Load timing turns out to be the axis that genuinely separates these, and once it's visible the categories fall out on their own:

  • agents / instructions / skills → become context
  • hooks / workflows → execute, and never enter the context window
  • plugins → packaging for the above

That grouping was implicit before. It's now readable straight off the table.

Added the missing hooks and workflows rows, closing the gap with the tagline.

Fixed three accuracy problems in the existing rows:

  • Skills previously implied the whole file is loaded on demand. It isn't — only the frontmatter description stays resident, and the body and bundled assets load when the agent opens the file. This matters practically: the description is doing all the routing work, which is why AGENTS.md requires it to be 10–1024 characters.
  • Agents said "pinned model." model is optional per AGENTS.md (strongly recommended, not required), and many agents in the collection omit it.
  • Plugins said "Once, at install time," which conflated the package arriving with its contents reaching the model. The row now defers to the rules above it.

Notes

  • The Browse links for hooks and workflows point at the repo directories rather than the website. I couldn't find awesome-copilot.github.com/hooks or /workflows pages referenced anywhere in the repo or under website/, so I didn't want to invent URLs. Happy to swap them in if those pages exist.
  • This section of README.md is hand-authored. eng/update-readme.mjs only splices between the ## 🌟 Featured Plugins and ## MCP Server markers, neither of which is present here — so the build doesn't manage this table.
  • Ran npm run build; it produced no changes beyond this edit.
  • eng/fix-line-endings.sh errors on macOS (sed: invalid command code) because it uses GNU sed -i syntax. It made no changes. Verified README.md is LF-only independently. Flagging as a possible separate issue.

The "What's in this repo" table described every resource type as a
generic piece of configuration, which gave newcomers no way to tell
them apart or decide which one to reach for.

- Add a "When it loads" column. Load timing is the axis that actually
  separates these: agents/instructions/skills are context, hooks and
  workflows are execution, plugins are packaging.
- Add the missing hooks and workflows rows. The tagline promises six
  resource types; the table only listed four.
- Describe skills as loading in two stages, since only the frontmatter
  description is resident until the agent opens the file.
- Correct the agents row: model is optional, not pinned.
- Correct the plugins row: install time is when the package arrives,
  not when its contents reach the model.
Copilot AI review requested due to automatic review settings August 2, 2026 14:29
@horner
horner requested a review from aaronpowell as a code owner August 2, 2026 14:29
@github-actions

github-actions Bot commented Aug 2, 2026

Copy link
Copy Markdown
Contributor

🟡 Contributor Reputation Check: MEDIUM risk

Check Risk
Profile MEDIUM
Credential audit NONE

Maintainers: please review this contributor before merging.
See the workflow run for full details.
Automated check powered by AGT.

@github-actions github-actions Bot added the needs-review:MEDIUM Contributor reputation check flagged MEDIUM risk label Aug 2, 2026

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Clarifies how repository resource types differ by runtime behavior.

Changes:

  • Adds load-timing guidance for each resource type.
  • Adds hooks and workflows to the comparison table.
  • Refines existing resource descriptions.

Comment thread README.md Outdated
Comment thread README.md
|----------|-------------|---------------|--------|
| 🤖 [Agents](docs/README.agents.md) | A specialist persona: system prompt + allowed tools/MCP servers + an optional pinned model | When **you** pick it for a conversation | [All agents →](https://awesome-copilot.github.com/agents) |
| 📋 [Instructions](docs/README.instructions.md) | Coding standards and repo conventions — prose only, no tools | **Automatically**, when a file matches its `applyTo` pattern | [All instructions →](https://awesome-copilot.github.com/instructions) |
| 🎯 [Skills](docs/README.skills.md) | Task procedures plus bundled scripts, templates, and reference data | **In two stages** — only the frontmatter `description` sits in context; the body and assets load when the agent opens the file | [All skills →](https://awesome-copilot.github.com/skills) |
Comment thread README.md
| 🤖 [Agents](docs/README.agents.md) | A specialist persona: system prompt + allowed tools/MCP servers + an optional pinned model | When **you** pick it for a conversation | [All agents →](https://awesome-copilot.github.com/agents) |
| 📋 [Instructions](docs/README.instructions.md) | Coding standards and repo conventions — prose only, no tools | **Automatically**, when a file matches its `applyTo` pattern | [All instructions →](https://awesome-copilot.github.com/instructions) |
| 🎯 [Skills](docs/README.skills.md) | Task procedures plus bundled scripts, templates, and reference data | **In two stages** — only the frontmatter `description` sits in context; the body and assets load when the agent opens the file | [All skills →](https://awesome-copilot.github.com/skills) |
| 🪝 [Hooks](docs/README.hooks.md) | Shell commands wired to coding-agent lifecycle events like `sessionStart` or `preToolUse` | Never — hooks *run*, they don't enter the context window | [Browse hooks →](hooks/) |
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Copilot AI review requested due to automatic review settings August 3, 2026 13:54

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated no new comments.

Suppressed comments (2)

README.md:25

  • This collapses the skill specification's three disclosure levels into two. Discovery metadata includes both name and description; the SKILL.md body loads when the skill is activated, while bundled resources are loaded separately only as needed—not all when the agent opens SKILL.md. Since this row is intended to explain routing and context cost, please describe those stages separately.
| 🎯 [Skills](docs/README.skills.md) | Task procedures plus bundled scripts, templates, and reference data | **In two stages** — only the frontmatter `description` sits in context; the body and assets load when the agent opens the file | [All skills →](https://awesome-copilot.github.com/skills) |

README.md:28

  • The repository's plugin contract is broader than agents and skills: AGENTS.md:12,126 explicitly includes commands, and existing manifests can also register integrations such as mcpServers (plugins/awesome-copilot/.github/plugin/plugin.json:17-25). Restricting both cells to agents/skills leaves these bundled components out of the taxonomy and gives them no activation rule.
| 🔌 [Plugins](docs/README.plugins.md) | Installable bundles of related agents and skills — packaging, not a new capability | **At install time only** — the bundled agents and skills then load by their own rules above | [All plugins →](https://awesome-copilot.github.com/plugins) |

Comment thread README.md
| 🎯 [Skills](docs/README.skills.md) | Task procedures plus bundled scripts, templates, and reference data | **In two stages** — only the frontmatter `description` sits in context; the body and assets load when the agent opens the file | [All skills →](https://awesome-copilot.github.com/skills) |
| 🪝 [Hooks](docs/README.hooks.md) | Shell commands wired to coding-agent lifecycle events like `sessionStart` or `preToolUse` | Never — hooks *run*, they don't enter the context window | [Browse hooks →](hooks/) |
| ⚙️ [Workflows](docs/README.workflows.md) | Agentic workflows that run Copilot in GitHub Actions on a schedule or repo event | Not in your editor — these execute in CI | [Browse workflows →](workflows/) |
| 🔌 [Plugins](docs/README.plugins.md) | Installable bundles of related agents and skills — packaging, not a new capability | **At install time only** — the bundled agents and skills then load by their own rules above | [All plugins →](https://awesome-copilot.github.com/plugins) |

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The "When it loads" here isn't accurate as a plugin itself isn't "loaded", it's the items within the plugin, the agents/skills/etc. that are loaded using the rules outlined above.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

needs-review:MEDIUM Contributor reputation check flagged MEDIUM risk

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants