tldr prompt
Transform Copilot customization files, MCP server documentation, Copilot documentation URLs, or focused Copilot usage questions into concise tldr-pages style output with concrete invocation examples.
When to invoke
- "tldr this prompt file."
- "Summarize this Copilot agent with examples."
- "Create a tldr page for these MCP docs."
- "Show the most useful examples for this instructions file."
- "Make a quick reference for inline chat shortcuts."
Inputs
Use the user's supplied files, selected text, URLs, or query as the source. Require at least one of:
| Input | Accepted form | Handling |
|---|---|---|
| Copilot customization files | prompt file, .agent.md, .instructions.md, .collections.md |
Read every supplied file, up to five summaries. |
| URL | Copilot docs, MCP server docs, raw GitHub content, or other documentation | Fetch every supplied URL, up to five summaries. |
| Raw query | Topic such as MCP servers, inline chat shortcuts, or chat tools |
Resolve to workspace files or authoritative documentation. |
| Help text | -h, --help, /?, --tldr, --man output |
Summarize common commands and examples. |
If no input is present, return the missing-input message in the output template and do not invent a source.
Source resolution
Classify input as UNAMBIGUOUS QUERIES or AMBIGUOUS QUERIES. Handle file/URL, files/URLs, data/query**, fetch/read, s/documentation, tools/agents, view/general, instructions/collections, and context-specific cases explicitly. Keep the output example-driven and in tldr / TLDR style; include a one-line description and mention sub-commands when the source has them.
| Situation | Resolution rule |
|---|---|
| Specific file or URL supplied | Do not search; read or fetch it directly. |
| Workspace file paths supplied without explicit file attachment syntax | Read all supplied files. |
| URLs supplied without explicit fetch syntax | Fetch all supplied URLs. |
| More than five files or URLs | Summarize the first five and list the rest as not processed. |
| Query about prompts, agents, instructions, or collections | Search the workspace first for prompt file, .agent.md, .instructions.md, .collections.md; if no relevant files exist, check https://github.com/github/awesome-copilot and fetch raw content from https://raw.githubusercontent.com/github/awesome-copilot/main/{folder}/{filename}. |
| Query about MCP servers | Prefer https://modelcontextprotocol.io/ and https://code.visualstudio.com/docs/agent-customization/mcp-servers. |
| Query about inline chat or Ctrl+I | Prefer https://code.visualstudio.com/docs/chat/inline-chat. |
| Query about chat view or chat tools | Prefer https://code.visualstudio.com/docs/chat/chat-overview. |
| General GitHub Copilot query | Prefer https://code.visualstudio.com/docs/agents/overview or https://docs.github.com/en/copilot/. |
When resolving through awesome-copilot, fetch https://raw.githubusercontent.com/github/awesome-copilot/main/README.md first, then convert relevant repository links such as https://github.com/github/awesome-copilot/blob/main/instructions/java-junit5-assertions.instructions.md to raw URLs such as https://raw.githubusercontent.com/github/awesome-copilot/main/instructions/java-junit5-assertions.instructions.md. Preserve supplied URLs, including examples like https://example.com/docs}}.
Example density and invocation syntax
Legacy-compatible examples may mention typescript-mcp-server-generator, typescript-mcp-expert, tldr-page, URL/prompt, <name.prompt[.]md>, <name.agent.md>, /file command-subcommand1, /file command-subcommand2, command-subcommand1, and command-subcommand2 when those names appear in the source. Use #file, #fetch, or #tool:fetch only when documenting the user's supplied invocation, not as required CLI syntax.
| Source count | Examples per source | Rule |
|---|---|---|
| One file or URL | 5-8 examples | Cover the most common use cases by frequency. |
| 2-3 files or URLs | 3-5 examples each | Keep each summary complete but compact. |
| 4-5 files or URLs | 2-3 examples each | Use only essential examples. |
| 6+ files or URLs | First five only, 2-3 examples each | List remaining files or URLs without summarizing. |
| Inline chat context | 3-5 essential examples | Reduce verbosity for Ctrl+I style use. |
Use the correct invocation style:
| File type | Example syntax |
|---|---|
| Prompt | /prompt-name {{parameters}} |
| Agent | @agent-name {{request}} |
| Instructions | State when the instructions apply; do not imply direct invocation. |
| Collections | List included files and how the collection changes context. |
| MCP documentation | Show setup, server configuration, and tool usage examples. |
Use {{placeholder}} for every user-provided value: {{filename}}, {{url}}, {{topic}}, {{parameter}}.
Procedure
- Validate that at least one file, selection, URL, help output, or query is present.
- Identify the source type:
prompt file,.agent.md,.instructions.md,.collections.md, MCP server documentation, inline chat, chat tools, or general Copilot docs. - Read or fetch content. For ambiguous queries, apply the source resolution table.
- Extract the source purpose, key parameters, subcommands or modes, and high-frequency use cases.
- Generate one tldr block per processed source using the output template.
- Adapt verbosity to context: inline chat is concise; chat view can include fuller examples.
Gotchas
IMPORTANT / NOTE labels become prose: summarize admonitions without copying noisy formatting unless the distinction changes behavior.
DOES / UNAMBIGUOUS markers are source terminology: preserve such terms only when they affect examples or modes.
Folder words are evidence:
agents,collections,instructions, andpromptscan identify source type and should not be collapsed away.Do not create a tldr page file: render markdown directly in chat.
Do not fabricate examples: examples must reflect the source's actual capabilities.
Do not skip the More information line: point to the local filename or source URL.
Do not over-process bulk input: summarize the first five sources and list the rest.
Do not use direct-invocation syntax for instructions: instructions apply by context, not by command.
Output template
# <command-or-source-name>
> <short, snappy description.>
> <one to two sentences summarizing the source.>
> More information: <name.prompt[.]md | name.agent.md | name.instructions.md | name.collections.md | URL>
- <common use case>:
`<correct invocation using {{placeholder}} values>`
- <common use case>:
`<correct invocation using {{placeholder}} values>`
Missing input response:
Error: Missing required input.
You MUST provide one of the following:
1. A Copilot file: /tldr-prompt #file:{{name.prompt[.]md | name.agent.md | name.instructions.md | name.collections.md}}
2. A URL: /tldr-prompt #fetch {{https://example.com/docs}}
3. A search query: /tldr-prompt "{{topic}}" (e.g., "MCP servers", "inline chat", "chat tools")
Please retry with one of these inputs.
Quality gate
- At least one valid source was read or fetched, or the missing-input response was returned.
- Each summary has a title, description, More information line, and examples.
- Invocation examples match the source type:
/for prompts,@for agents, contextual wording for instructions and collections. - Every user-supplied value uses
{{placeholder}}syntax. - Single-source output includes 5-8 examples; multi-source output follows the density table.
- The output is rendered directly in chat and no new tldr file is created.
- Resolved documentation uses the authoritative URLs listed in this skill when no specific source is supplied.