fix(mcp): return owner's own private prompts from get_prompt + accept category name/slug#1215
Open
fawadafr wants to merge 2 commits into
Open
fix(mcp): return owner's own private prompts from get_prompt + accept category name/slug#1215fawadafr wants to merge 2 commits into
fawadafr wants to merge 2 commits into
Conversation
Previously the MCP save tools matched `category` by exact slug only and
silently dropped any unrecognized value, so prompts were created with no
category and no feedback. Clients also had no way to discover valid slugs:
search_prompts exposed only the category name, and there was no listing tool.
- Add `list_categories` tool returning {name, slug, description, promptCount}
- Resolve category by slug, name, or slugified-name (forgiving match)
- Return a `warning` when a category is provided but unmatched, instead of
failing silently — naming the valid slugs
- Expose `categorySlug` in search_prompts and the save responses for round-trip
- Unit-test the matcher (findCategoryMatch) for the resolution cases
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SxPKXfHMaqgtmitAQC1Zjx
get_prompt hard-filtered `isPrivate: false`, so the authenticated owner got "Prompt not found" for their own private prompts — while search_prompts and get_skill already include the owner's private entries. This blocked reading a private prompt's full content (e.g. to score it against an eval). Extract buildPromptVisibilityFilter() (public OR owner's-own-private) and use it in both get_prompt and get_skill, replacing the duplicated inline blocks. Add unit tests for the helper incl. a regression check. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DL6AEgE5FypVamGVTZhS8a
📝 WalkthroughWalkthroughAdds MCP Category Support
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Possibly related PRs
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@src/pages/api/mcp.ts`:
- Around line 1424-1445: The list_categories response is leaking hidden prompt
volume through the unfiltered promptCount value. In the category query and
mapping inside the list_categories handler in mcp.ts, either remove promptCount
from the returned category payload or compute it using the same
caller-visibility constraints used by the prompt lookup tools so only visible
prompts are counted.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: CHILL
Plan: Pro
Run ID: f5570d71-a251-498e-891e-61513e2c9013
📒 Files selected for processing (2)
src/__tests__/api/mcp-handler.test.tssrc/pages/api/mcp.ts
Comment on lines
+1424
to
+1445
| const categories = await db.category.findMany({ | ||
| orderBy: [{ order: "asc" }, { name: "asc" }], | ||
| select: { | ||
| name: true, | ||
| slug: true, | ||
| description: true, | ||
| _count: { select: { prompts: true } }, | ||
| }, | ||
| }); | ||
|
|
||
| return { | ||
| content: [ | ||
| { | ||
| type: "text" as const, | ||
| text: JSON.stringify({ | ||
| count: categories.length, | ||
| categories: categories.map((c) => ({ | ||
| name: c.name, | ||
| slug: c.slug, | ||
| description: c.description || null, | ||
| promptCount: c._count.prompts, | ||
| })), |
There was a problem hiding this comment.
🔒 Security & Privacy | 🟠 Major | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Verify the Category -> Prompt relation and existing count/filter test coverage.
fd -a 'schema.prisma$' . -x sed -n '/model Category/,/}/p' {}
fd -a 'schema.prisma$' . -x sed -n '/model Prompt/,/}/p' {}
rg -n -C3 'list_categories|_count:\s*\{\s*select:\s*\{\s*prompts|promptCount|buildPromptVisibilityFilter' src/pages/api/mcp.ts src/__tests__/api/mcp-handler.test.tsRepository: f/prompts.chat
Length of output: 12962
🏁 Script executed:
#!/bin/bash
set -euo pipefail
# Check whether this repo already uses relation-count filtering elsewhere.
rg -n '_count:\s*\{\s*select:\s*\{\s*[^}]+where:' srcRepository: f/prompts.chat
Length of output: 152
Filter promptCount by caller visibility, or drop it. list_categories currently counts every related prompt, so callers can infer hidden prompt volume from category stats. Use the same visibility filter as the prompt lookup tools.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@src/pages/api/mcp.ts` around lines 1424 - 1445, The list_categories response
is leaking hidden prompt volume through the unfiltered promptCount value. In the
category query and mapping inside the list_categories handler in mcp.ts, either
remove promptCount from the returned category payload or compute it using the
same caller-visibility constraints used by the prompt lookup tools so only
visible prompts are counted.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Two small, self-contained fixes to the MCP server (
src/pages/api/mcp.ts), each with tests.1.
get_prompthides owners' own private promptsget_prompthard-filteredisPrivate: false, so an authenticated owner got "Prompt not found" when fetching their own private prompt — even thoughget_skillandsearch_promptsalready surface owner-private entries. This was an inconsistency between the two read paths.Extracted a shared
buildPromptVisibilityFilter(authenticatedUser)→isPrivate:false OR (isPrivate:true AND authorId === me)for authenticated callers, public-only otherwise — and applied it to bothget_promptandget_skillso they stay in sync.2.
save_prompt/save_skillsilently drop unrecognized categoriesCategory was resolved by exact slug only (
findUnique({ where: { slug } })); a category name, a differently-cased slug, orcode_reviewvscode-reviewwas silently ignored, creating an uncategorized prompt with no feedback.findCategoryMatch()/resolveCategory()match by slug → name → slugified-name (all case-insensitive).warninglisting valid slugs, instead of silently dropping the value.list_categoriestool so clients can discover valid slugs before saving.categorySlugtosearch_promptsand to thesave_*success payloads so a category can be round-tripped.Tests
src/__tests__/api/mcp-handler.test.tsadds unit coverage forfindCategoryMatch(slug / name / slugify / whitespace / no-match) andbuildPromptVisibilityFilter(including a regression test that an owner's filter does not hard-exclude their private prompts). All 17 tests pass.No schema changes, no new dependencies, no breaking changes to existing tool signatures (the
categoryarg now also accepts a name; previously-valid slugs behave identically).🤖 Generated with Claude Code
Implemented MCP server fixes around prompt visibility and category handling.
get_promptso authenticated owners can access their own private prompts, while unauthenticated users still only see public prompts.get_promptandget_skill.save_promptandsave_skillto resolve by slug, name, or slugified name in a case-insensitive way.list_categoriesfor client-side category lookup.categorySluginsearch_promptsand save success responses so category values can round-trip.