1.7 KiB
1.7 KiB
| title | state | created_at | updated_at | assignee | readiness | risk_flags | queued_by | queued_at | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MCP: expose resources and prompts as explicit tool operations | closed | 2026-06-20T05:30:04Z | 2026-06-20T10:05:16Z | null | implementation_ready |
|
workspace-panel | 2026-06-20T05:58:57Z |
Background
MCP resources and prompts must not become hidden context injection. They should be exposed as explicit Yoi tool operations whose results are recorded through ordinary Tool result/history paths.
Requirements
- Expose MCP resources/prompts as explicit namespaced Yoi tool operations:
resources/list,resources/read,prompts/list, andprompts/get. - Treat returned content/templates as untrusted tool result data.
- Do not inject resource/prompt content directly into context outside history/tool result.
- Bound result sizes and rich/embedded content serialization.
- Handle pagination/list bounds where applicable.
- Diagnostics identify server/resource/prompt operation without leaking secrets.
Acceptance criteria
resources/listandresources/readcan be invoked as explicit tools.prompts/listandprompts/getcan be invoked as explicit tools.- Results are ordinary Tool results and history records.
- No hidden context injection path is introduced.
- Oversize/rich content is bounded.
- Tests cover list/read/get happy paths, untrusted content, bounds, and no hidden injection.
Non-goals
- MCP tool execution itself.
- list_changed notification refresh.
- Sampling/elicitation.
Related work
- Depends on
00001KVHR3WRY. - Related to
00001KVHR3WSDfor result serialization policy. - Objective:
00001KTR80WMN.