Describe the feature or problem you'd like to solve
AI agents using the MCP server create manual markdown links when they could use GitHub's built-in autolinked references instead. GitHub automatically renders patterns like #123, @username, and owner/repo#123 as clickable links across issues, pull requests, comments, and discussions—but agents lack this knowledge in their system instructions.
For complete reference and examples, see:
https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/autolinked-references-and-urls
Proposed solution
Add guidance to the base context instructions (via generateContextToolsetInstructions()) so all agents learn about autolinked references before using any content-creation tools. This ensures consistent formatting across all GitHub interactions without duplicating guidance in multiple toolsets.
Benefits:
- Universal coverage: Applies to all content-creation tools (issues, PRs, comments, discussions)
- Single source of truth: One place to maintain, not scattered across toolsets
- Early discovery: Agents see this guidance at initialization, before using any specific tool
- No duplication: Leverage official docs instead of repeating content
Implementation:
Add a section to pkg/github/toolset_instructions.go in generateContextToolsetInstructions():
func generateContextToolsetInstructions(_ *inventory.Inventory) string {
return `Always call 'get_me' first to understand current user permissions and context.
### GitHub Autolinked References
When creating issues, pull requests, comments, or discussions, use GitHub's autolinked reference syntax—issue numbers (#123), user mentions (@example-username), and cross-repo references (example-owner/example-repo#123)—so GitHub renders them as links automatically.
See the official docs for complete details: https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/autolinked-references-and-urls.md`
}This approach follows the existing pattern of github_support_docs_search, which references official documentation rather than duplicating content.
Additional context
This is a lightweight enhancement that integrates seamlessly into the existing base context instructions. No new tools or infrastructure required—just a brief instruction pointing agents to official GitHub docs.
Future enhancement:
If agents need to proactively request formatting guidance (e.g., "How should I format this reference?"), consider creating a dedicated resource or skill in the MCP server. This would allow agents to ask for formatting help on-demand while still having the baseline guidance available at initialization. This can be explored as a follow-up if needed.
Describe the feature or problem you'd like to solve
AI agents using the MCP server create manual markdown links when they could use GitHub's built-in autolinked references instead. GitHub automatically renders patterns like
#123,@username, andowner/repo#123as clickable links across issues, pull requests, comments, and discussions—but agents lack this knowledge in their system instructions.For complete reference and examples, see:
https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/autolinked-references-and-urls
Proposed solution
Add guidance to the base context instructions (via
generateContextToolsetInstructions()) so all agents learn about autolinked references before using any content-creation tools. This ensures consistent formatting across all GitHub interactions without duplicating guidance in multiple toolsets.Benefits:
Implementation:
Add a section to
pkg/github/toolset_instructions.goingenerateContextToolsetInstructions():This approach follows the existing pattern of
github_support_docs_search, which references official documentation rather than duplicating content.Additional context
This is a lightweight enhancement that integrates seamlessly into the existing base context instructions. No new tools or infrastructure required—just a brief instruction pointing agents to official GitHub docs.
Future enhancement:
If agents need to proactively request formatting guidance (e.g., "How should I format this reference?"), consider creating a dedicated resource or skill in the MCP server. This would allow agents to ask for formatting help on-demand while still having the baseline guidance available at initialization. This can be explored as a follow-up if needed.