fix(mcp): keep boundary suggestions source-free

[JSONbored]

Motivation

• Prevent leakage of private local source by ensuring the MCP gittensory_suggest_boundary_tests tool no longer accepts raw unified-diff patch text at the remote boundary.
• Replace server-side scanning of raw patches with a source-free metadata contract so agents compute any patch-derived artifacts locally and only submit precomputed, non-sensitive metadata.

Description

• Change the MCP tool input schema for gittensory_suggest_boundary_tests to remove patch payloads and accept only changedFiles (paths) plus an optional, strict boundaryTouches array containing only path and kind (no source snippets), and cap touches to max(20). (src/mcp/server.ts)
• Update tool registration and description to document the new source-free contract and to filter supplied boundaryTouches to the declared changedFiles. (src/mcp/server.ts)
• Remove source snippets from BoundaryTouch and update detectBoundaryTouches to return only path + kind; make buildBoundaryTestGenerationFinding accept either precomputed touches or legacy files input so local callers can still scan and then submit safe metadata. (src/signals/boundary-test-generation.ts)
• Add/adjust unit tests to cover the regression: MCP now rejects raw patch input, accepts precomputed touches, ignores touches for files not declared in changedFiles, and verifies that returned touches contain no source snippets. (test/unit/mcp-suggest-boundary-tests.test.ts, test/unit/boundary-test-generation.test.ts)

Testing

• Ran type checking with npm run typecheck and it passed.
• Ran targeted unit suites with npx vitest run test/unit/boundary-test-generation.test.ts test/unit/mcp-suggest-boundary-tests.test.ts and all tests passed.
• Ran git diff --check which passed.
• Attempted full local gate via npm run test:ci; execution reached test:coverage but the run was interrupted due to an unrelated recursive RangeError: Maximum call stack size exceeded observed in an existing test/unit/queue.test.ts test, so the full CI gate was not completed here.
• npm audit --audit-level=moderate was attempted but the registry audit endpoint returned 403 Forbidden in this environment, so the dependency-review step could not be validated here.

---

Codex Task

[superagent-security]

Superagent didn't find any vulnerabilities or security issues in this PR.

[codecov]

Codecov Report

❌ Patch coverage is 80.00000% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 93.66%. Comparing base (1ce4364) to head (d284aba).
⚠️ Report is 15 commits behind head on main.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
src/signals/boundary-test-generation.ts	50.00%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3955      +/-   ##
==========================================
- Coverage   93.66%   93.66%   -0.01%     
==========================================
  Files         372      372              
  Lines       34856    34856              
  Branches    12743    12744       +1     
==========================================
- Hits        32649    32648       -1     
  Misses       1588     1588              
- Partials      619      620       +1     

Files with missing lines	Coverage Δ
src/mcp/server.ts	95.83% <100.00%> (+<0.01%)	⬆️
src/signals/boundary-test-generation.ts	97.14% <50.00%> (-2.86%)	⬇️

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[loopover-orb]

Caution

🟥🟥🟥🟥🟥🟥🟥🟥🟥🟥🟥🟥

🛑 Gittensory review result - fixes required

_{Review updated: 2026-07-07 09:14:17 UTC}

4 files · 1 AI reviewer · 2 blockers · readiness 93/100 · CI failing · unstable

🛑 Suggested Action - Manual Review

• No linked issue detected — If this PR is intended to solve an issue, link it explicitly in the PR body.
• Maintainer requires a linked issue — Link the relevant issue (for example Closes #123) before opening the PR.

Review summary
This PR closes a real boundary leak: the MCP `gittensory_suggest_boundary_tests` tool previously accepted raw unified-diff `patch` text over the remote boundary, and now only accepts file paths plus a strict, capped `boundaryTouches` metadata array (path + pattern kind, no snippets). The server-side filter in `suggestBoundaryTests` (src/mcp/server.ts) correctly restricts supplied touches to the declared `changedFiles` set before building the finding/spec, and `BoundaryTouch`/`detectBoundaryTouches`/`buildBoundaryTestGenerationFinding` in src/signals/boundary-test-generation.ts were updated consistently to drop `snippet` while preserving a `files`-based legacy path for other in-repo callers. Tests were added for the new rejection (`patch` now errors), the touch-filtering behavior, and the absence of `snippet` in output, and the changes are internally consistent with no schema/migration surface. The one open gap is coverage: codecov/patch sits at 80% vs the 99% target, so some new lines/branches introduced here (e.g. the `files ?? []` fallback branch of `buildBoundaryTestGenerationFinding`, or interactions between `boundaryTouches` omitted vs empty) likely lack direct tests.

Nits — 5 non-blocking

• src/mcp/server.ts:808 hardcodes `.max(20)` for `boundaryTouches` even though `MAX_TOUCHES = 20` already exists in src/signals/boundary-test-generation.ts — export and reuse that constant so the schema cap can't silently drift out of sync with the signal's own cap.
• codecov/patch is failing at 80% vs the 99% target for this diff — worth adding a test for the `input.files ?? []` / no-files-no-touches path in `buildBoundaryTestGenerationFinding` (src/signals/boundary-test-generation.ts) since that arm isn't obviously exercised by the new tests.
• Confirm `detectBoundaryTouches` and the `files`-based branch of `buildBoundaryTestGenerationFinding` still have a real in-repo caller now that src/mcp/server.ts no longer imports `detectBoundaryTouches` directly — if nothing else calls the legacy `files` path in production, it's now dead code kept alive only by tests.
• The PR description doesn't cite a maintainer-authorized issue beyond the pre-existing `Boundary-safe test generation #1972` reference in code comments — worth confirming this security fix is explicitly linked to an eligible open issue per repo convention.
• Export `MAX_TOUCHES` from src/signals/boundary-test-generation.ts and reference it in the `boundaryTouches` zod schema's `.max()` in src/mcp/server.ts instead of the literal `20`.

Why this is blocked

• No linked issue detected — If this PR is intended to solve an issue, link it explicitly in the PR body.
• Maintainer requires a linked issue — Link the relevant issue (for example Closes #123) before opening the PR.

CI checks failing

• codecov/patch — 80.00% of diff hit (target 99.00%)

Signal	Result	Evidence
Code review	❌ 2 blockers	1 reviewer
Linked issue	⚠️ Missing	No linked issue or no-issue rationale found.
Related work	✅ No active overlap found	No same-issue or scoped active PR overlap found.
Change scope	✅ 20/20	Low review scope from cached public metadata (no linked issue context).
Validation posture	✅ 25/25	PR body includes validation/test evidence.
Contributor workload	✅ 10/10	Author activity: 51 registered-repo PR(s), 43 merged, 569 issue(s).
Contributor context	✅ Confirmed Gittensor contributor	JSONbored; Gittensor profile; 51 PR(s), 569 issue(s).
Gate result	❌ Blocking	Repo-configured hard blocker found.

Review context

• Author: JSONbored
• Role context: owner (maintainer lane)
• Public audience mode: oss maintainer
• Lane context: Repository is configured for direct PR review.
• Public profile languages: Python, TypeScript, JavaScript, Ruby, Go, Kotlin, MDX, Shell
• Official Gittensor activity: 51 PR(s), 569 issue(s).
• PR-specific overlap: none found.

Contributor next steps

• Treat this as maintainer-lane context rather than normal contributor-lane activity.
• Explain no-issue PR.
• Link the issue being solved, or explicitly explain why this is a no-issue PR.

Signal definitions

• Related work = same linked issue, overlapping active PRs, or title/path similarity.
• Change scope = cached public metadata such as size labels, draft state, and review-burden hints.
• Validation posture = whether the PR provides enough public validation/test evidence for maintainer review.
• Contributor workload = public contributor activity and cleanup pressure, not a repo-wide quality failure.
• Contributor context = public GitHub/Gittensor identity context; non-Gittensor status is not a blocker.

_{🟩 Safe / merged · 🟦 Advisory · 🟨 Held for review · 🟥 Blocked / closed}

---

💰 Earn for open-source contributions like this. Gittensor lets GitHub contributors earn for the work they already do — register to start earning →.

Checked by Gittensory, a quiet PR intelligence layer for OSS maintainers.

• Re-run Gittensory review