Repository navigation
Conversation
First Principles Review (Opus 5.5) — ⏭️ skippedRevision |
UX Review (Opus 5.5) — ⏭️ skippedRevision |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Nothing blocks: this PR only adds a draft RFC and its index row, and it changes no code. It names the one boundary move (Stage 2b) and holds it on Q1. Notes
[DESIGN-REVIEWED] 9198c0e Description read[DESCRIPTION-READ] fbb5184033beadf684a5f5239fd8569a3cf21bf54391f0d3a09bc67f8d4876ac That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
Opus 5.5 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
c325aa3 to
4edd338
Compare
Code change approver recommendation: A person should decideWhy: the decider estimates a 12% chance that this change introduces a defect, within the 12% limit for approving on its own, but it is held at tier 2: design-doc-gate: D. Docs-only draft RFC plus one index row; it changes no code. Its claims about today's code check out: probe_server refuses secret:// env, record_verdicts saves the last error and ok clears it, _annotate_quarantine copies only probeFailures and probeFailing, both write routes are owner-gated for browser callers, and the featured list is bundled with no order. The one trust-boundary move (stage 2b) is named and blocked on open question Q1, and stage 2a ships without it if the answer is no. Design gaps (stage 2b reading keys for servers someone else registered, vault-name collisions, nothing recorded when quarantine is off) are for the maintainer to settle before implementation and do not make this draft unsound to merge. Tier: 2, held for a person: design-doc-gate: D. An automated recommendation, not a review: a maintainer still decides. It approves on its own only when the estimated chance of a defect is at most 12% and the change has at most 400 changed lines. Decider v3c (217fe607) · commit |
4edd338 to
086744e
Compare
|
This changes behaviour users already rely on, so it needs a design sign-off from a maintainer before it can merge. A maintainer will post the decision here. Review result: this review found no blocking findings on this head. The PR adds a draft RFC and one index row. It changes no code. We checked the RFC's claims about today's code against main, and almost all of them hold:
Done well: each claim names a concrete symbol. Stage 2b, the one trust-boundary move, has its own open question and a fallback. For the security review:
Optional:
|
086744e to
9198c0e
Compare
Problem / Motivation
Goal: Propose, as a draft RFC, the three MCP Connections Center stages that follow the per-row Test button, so a maintainer can decide them before any code is written.
Stage 1, the per-row Test button on Customize → MCP Servers, shipped in #38055. On #38054 a maintainer asked for a request-for-change before the remaining stages are built. Stage 2 writes a secret from the browser and would let one probe resolve a
secret://value, which moves a trust boundary.Why it matters
A person adding an MCP server still edits a config file to set its key, and the Test button cannot check a server whose key is in the vault. Writing the design down first lets a maintainer accept, change or drop the boundary move before it is built.
Not a goal
What changed (motivation → approach → change)
Symptom: the stages after the Test button have no agreed design, and one of them moves a trust boundary. Cause: #38054 holds them as a request with no RFC. Change: adds
docs/request-for-change/rfc-mcp-connections-center.md(statusdraft) and its index row indocs/request-for-change/README.md. Every claim about today's code was checked at3cc96a36de. Stage 3 reuses the last error the quarantine store already saves (record_verdictsinsrc/kiro_crew/mcp_quarantine.py); it adds no second copy.Backwards compatibility
Compatible: documentation only. No code, config, route or stored value changes.
Removes nothing: no route, field or value is taken away.
Tests
No code changes.
scripts/docs-lint.shpasses, andtest/test_pr_quality_gates.pyandtest/test_frontmatter.pypass (358 tests).Manual verification
N/A — documentation only: the change is a design document and an index row.
Real-machine verification
N/A -- the diff runs nothing: it adds one Markdown design document and one index row.
Screenshots / video
N/A — no UI change.
Related Issues
Part of #38054
Sibling sweep
Swept:
ls docs/request-for-change | grep -i mcpand opendocs(rfc)pull requests.Sibling:
docs/request-for-change/rfc-connections-mcp-layout.md-- not affected -- it records where MCP Servers lives, not credentials or errorsSibling:
docs/request-for-change/rfc-owner-written-mcp-auto-approve.md-- not affected -- it covers tool approvals, not server keysSibling: #33754, #24872, #21975 (open MCP RFC pull requests) -- not affected -- none covers server credentials, last errors or the featured list
Pattern harvest
Rule candidate: none.
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)Contribution License Agreement
Contributed under the project's Apache-2.0 license.