Update to Markdown by mdedit.ai
Snapshot Oct 24, 2026 · 12:55 UTC · version 0.2.2
Collection source: downloaded plugin package. These snapshots do not have a confirmed matching collection source. Differences in file lists alone do not establish changes to the package.
Instructions updated for find-mdedit-document
Skill instructions — Now says: ## Authentication Use the host's browser OAuth connection flow when authentication is needed. The hosted server discovers Clerk and supports dynamic client...
Observed in instructions or declared skills. Runtime behavior has not been tested.
Up to 3 plugins free · no card required.
All fields and original values.
Skill instructions
Sep 30, 2026 · 22:55 UTC
1. Resolve the workspace with `list_workspaces` unless a `workspaceId` from an earlier mdedit tool result is already established. 2. Call `list_articles` for the resolved workspace. 4. Call `read_article` with the returned `workspaceId` ...
Oct 24, 2026 · 12:55 UTC
## Authentication Use the host's browser OAuth connection flow when authentication is needed. The hosted server discovers Clerk and supports dynamic client registration (DCR) and Client ID Metadata Documents (CIMD); a pre-registered host...
Compare saved observations
Download comparison JSONSKILL.md text changes
− removed · + added
--- before +++ after @@ -6,17 +6,24 @@ # Find and read a document in mdedit +## Authentication + +Use the host's browser OAuth connection flow when authentication is needed. The hosted server discovers Clerk and supports dynamic client registration (DCR) and Client ID Metadata Documents (CIMD); a pre-registered host client or API key is not required for an OAuth-capable host. Ask the user to sign in and approve the workspace access shown by mdedit.ai. Available workspaces are selected by default, with read/write access when the request includes write scopes; the user can narrow access before approving. + +Do not ask the user to paste an access token, client secret, authorization code, or API key into chat. Use an existing scoped API-key connection only when the host cannot perform OAuth or the user explicitly chooses it. If a tool reports missing permissions, reconnect through the host to request the needed scopes/workspaces instead of repeatedly retrying the same denied call. + + Resolve a real saved document before describing its content. -1. Resolve the workspace with `list_workspaces` unless a `workspaceId` from an earlier mdedit tool result is already established. +1. Resolve the workspace with `list_workspaces`. - If several workspaces could match, ask the user to choose by returned name. - If no workspace matches, stop without guessing. -2. Call `list_articles` for the resolved workspace. +2. Call `list_articles`. If exactly one workspace was returned, omit the workspace selector. If the user chose among several, pass `workspaceName`. When the user gives a title or keywords, pass them as `query`. Work from one bounded page at a time; do not retrieve every page merely to reproduce a long document list in chat. 3. Match the user's title or description against returned document metadata. - If there is one clear match, continue. - If several documents plausibly match, present only their returned titles and ask the user to choose. - If there is no match, say so and stop. -4. Call `read_article` with the returned `workspaceId` and `articleId`. -5. Answer from the returned saved `content`, `contentRevision`, and `contentHash`. Distinguish exact saved content from any summary or interpretation. +4. Call `read_article` with `articleTitle` and, when needed, `workspaceName`. Do not search local code or ask the user for routing identifiers. +5. For open, show, or preview requests, call `render_article` with `focus: preview`. Answer from the returned saved `content` when a textual answer is also needed. Distinguish exact saved content from any summary or interpretation. Keep `contentRevision`, `contentHash`, `workspaceId`, and `articleId` internal unless the user explicitly asks for diagnostic details. This workflow is read-only. Do not call create, edit, publish, or unpublish tools. Never infer document content, identifiers, revisions, publication status, or URLs from titles or conversation memory.
Full technical diff · 1 changed fields
changed /skill_md_contents
"---\nname: find-mdedit-document\ndescription: Find and read an existing Markdown Document stored in mdedit. Use when the user asks to locate, open, retrieve, quote, summarize, or inspect a saved mdedit document. Do not use for local files, web pages, or generic Markdown questions unrelated to mdedit storage.\ncompatibility: Requires the mdedit MCP server with workspace and article read tools.\n---\n\n# Find and read a document in mdedit\n\nResolve a real saved document before describing its content.\n\n1. Resolve the workspace with `list_workspaces` unless a `workspaceId` from an earlier mdedit tool result is already established.\n - If several workspaces could match, ask the user to choose by returned name.\n - If no workspace matches, stop without guessing.\n2. Call `list_articles` for the resolved workspace.\n3. Match the user's title or description against returned document metadata.\n - If there is one clear match, continue.\n - If several documents plausibly match, present only their returned titles and ask the user to choose.\n - If there is no match, say so and stop.\n4. Call `read_article` with the returned `workspaceId` and `articleId`.\n5. Answer from the returned saved `content`, `contentRevision`, and `contentHash`. Distinguish exact saved content from any summary or interpretation.\n\nThis workflow is read-only. Do not call create, edit, publish, or unpublish tools. Never infer document content, identifiers, revisions, publication status, or URLs from titles or conversation memory.\n"
"---\nname: find-mdedit-document\ndescription: Find and read an existing Markdown Document stored in mdedit. Use when the user asks to locate, open, retrieve, quote, summarize, or inspect a saved mdedit document. Do not use for local files, web pages, or generic Markdown questions unrelated to mdedit storage.\ncompatibility: Requires the mdedit MCP server with workspace and article read tools.\n---\n\n# Find and read a document in mdedit\n\n## Authentication\n\nUse the host's browser OAuth connection flow when authentication is needed. The hosted server discovers Clerk and supports dynamic client registration (DCR) and Client ID Metadata Documents (CIMD); a pre-registered host client or API key is not required for an OAuth-capable host. Ask the user to sign in and approve the workspace access shown by mdedit.ai. Available workspaces are selected by default, with read/write access when the request includes write scopes; the user can narrow access before approving.\n\nDo not ask the user to paste an access token, client secret, authorization code, or API key into chat. Use an existing scoped API-key connection only when the host cannot perform OAuth or the user explicitly chooses it. If a tool reports missing permissions, reconnect through the host to request the needed scopes/workspaces instead of repeatedly retrying the same denied call.\n\n\nResolve a real saved document before describing its content.\n\n1. Resolve the workspace with `list_workspaces`.\n - If several workspaces could match, ask the user to choose by returned name.\n - If no workspace matches, stop without guessing.\n2. Call `list_articles`. If exactly one workspace was returned, omit the workspace selector. If the user chose among several, pass `workspaceName`. When the user gives a title or keywords, pass them as `query`. Work from one bounded page at a time; do not retrieve every page merely to reproduce a long document list in chat.\n3. Match the user's title or description against returned document metadata.\n - If there is one clear match, continue.\n - If several documents plausibly match, present only their returned titles and ask the user to choose.\n - If there is no match, say so and stop.\n4. Call `read_article` with `articleTitle` and, when needed, `workspaceName`. Do not search local code or ask the user for routing identifiers.\n5. For open, show, or preview requests, call `render_article` with `focus: preview`. Answer from the returned saved `content` when a textual answer is also needed. Distinguish exact saved content from any summary or interpretation. Keep `contentRevision`, `contentHash`, `workspaceId`, and `articleId` internal unless the user explicitly asks for diagnostic details.\n\nThis workflow is read-only. Do not call create, edit, publish, or unpublish tools. Never infer document content, identifiers, revisions, publication status, or URLs from titles or conversation memory.\n"
Full snapshot data
{
"description": "Find and read an existing Markdown Document stored in mdedit. Use when the user asks to locate, open, retrieve, quote, summarize, or inspect a saved mdedit document. Do not use for local files, web pages, or generic Markdown questions unrelated to mdedit storage.",
"included_files": [
{
"relative_path": "agents/openai.yaml",
"size_in_bytes": 405
}
],
"name": "find-mdedit-document",
"skill_md_contents": "---\nname: find-mdedit-document\ndescription: Find and read an existing Markdown Document stored in mdedit. Use when the user asks to locate, open, retrieve, quote, summarize, or inspect a saved mdedit document. Do not use for local files, web pages, or generic Markdown questions unrelated to mdedit storage.\ncompatibility: Requires the mdedit MCP server with workspace and article read tools.\n---\n\n# Find and read a document in mdedit\n\n## Authentication\n\nUse the host's browser OAuth connection flow when authentication is needed. The hosted server discovers Clerk and supports dynamic client registration (DCR) and Client ID Metadata Documents (CIMD); a pre-registered host client or API key is not required for an OAuth-capable host. Ask the user to sign in and approve the workspace access shown by mdedit.ai. Available workspaces are selected by default, with read/write access when the request includes write scopes; the user can narrow access before approving.\n\nDo not ask the user to paste an access token, client secret, authorization code, or API key into chat. Use an existing scoped API-key connection only when the host cannot perform OAuth or the user explicitly chooses it. If a tool reports missing permissions, reconnect through the host to request the needed scopes/workspaces instead of repeatedly retrying the same denied call.\n\n\nResolve a real saved document before describing its content.\n\n1. Resolve the workspace with `list_workspaces`.\n - If several workspaces could match, ask the user to choose by returned name.\n - If no workspace matches, stop without guessing.\n2. Call `list_articles`. If exactly one workspace was returned, omit the workspace selector. If the user chose among several, pass `workspaceName`. When the user gives a title or keywords, pass them as `query`. Work from one bounded page at a time; do not retrieve every page merely to reproduce a long document list in chat.\n3. Match the user's title or description against returned document metadata.\n - If there is one clear match, continue.\n - If several documents plausibly match, present only their returned titles and ask the user to choose.\n - If there is no match, say so and stop.\n4. Call `read_article` with `articleTitle` and, when needed, `workspaceName`. Do not search local code or ask the user for routing identifiers.\n5. For open, show, or preview requests, call `render_article` with `focus: preview`. Answer from the returned saved `content` when a textual answer is also needed. Distinguish exact saved content from any summary or interpretation. Keep `contentRevision`, `contentHash`, `workspaceId`, and `articleId` internal unless the user explicitly asks for diagnostic details.\n\nThis workflow is read-only. Do not call create, edit, publish, or unpublish tools. Never infer document content, identifiers, revisions, publication status, or URLs from titles or conversation memory.\n"
}SHA-256 of public snapshot: c3a5bf78c13a2311ca1d4c1330bd1c0ec3abec5e93b81e401266bc6010e7e3c5