feat(app): mask secrets in API key and MCP install snippets with a shared RevealSnippet - #2988
Merged
Merged
Conversation
… snippets Introduce a shared RevealSnippet compound component that masks secrets in code snippets until revealed, while copy always yields the real value. Adopt it across the Team Settings API-key fields and the MCP install snippets (CLI command, deep-link JSON fallback, plain JSON). Masking is driven entirely by the secrets prop: omitted/empty renders a plain snippet, a bare string masks the whole value via defaultRedact, and [real, redacted] pairs mask specific substrings.
🦋 Changeset detectedLatest commit: 9f403bb The changes in this PR will be included in the next version bump. This PR includes changesets to release 3 packages
Not sure what this means? Click here to learn what changesets are. Click here if you're a maintainer who wants to add another changeset to this PR |
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
Contributor
🟡 Tier 3 — StandardIntroduces new logic, modifies core functionality, or touches areas with non-trivial risk. Why this tier:
Review process: Full human review — logic, architecture, edge cases. Stats
|
Contributor
Greptile SummaryThe PR adds a shared reveal-aware snippet component and adopts it for Team Settings credentials and MCP installation snippets.
Confidence Score: 5/5The PR appears safe to merge. No blocking failure remains.
|
| Filename | Overview |
|---|---|
| packages/app/src/components/RevealSnippet/RevealSnippet.tsx | Introduces the compound masking, reveal, input, and copy primitive; the previous key-rotation issue is fixed by resetting reveal state when the value changes. |
| packages/app/src/components/ClickStackOnboarding/McpInstallPanel.tsx | Passes the deployment access key through every visible MCP host snippet so credentials are masked consistently. |
| packages/app/src/components/ClickStackOnboarding/CopySnippet.tsx | Uses RevealSnippet for credential-bearing snippets while retaining the existing plain-snippet path. |
| packages/app/src/components/TeamSettings/ApiKeysSection.tsx | Replaces plaintext key controls with masked read-only fields containing reveal and copy actions. |
| packages/app/src/components/RevealSnippet/tests/RevealSnippet.test.tsx | Covers default masking, reveal behavior, real-value copying, nullable secrets, and re-masking after value changes. |
Reviews (3): Last reviewed commit: "Merge branch 'main' into brandon/reveal-..." | Re-trigger Greptile
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
Contributor
Deep Review✅ No critical issues found. The previously flagged P1 (reveal state surviving key rotation) is resolved in this branch — 🟡 P2 — recommended
🔵 P3 nitpicks (1)
Reviewers (7): correctness, testing, maintainability, project-standards, kieran-typescript, security, previous-comments. Testing gaps:
Notes: The Cursor/VS Code deep links carry the base64/URL-encoded key in the DOM |
Contributor
E2E Test Results✅ All tests passed • 307 passed • 1 skipped • 1191s
Tests ran across 4 shards in parallel. |
- Reset reveal state when the value changes (e.g. API key rotation) so a freshly issued credential is never shown without a fresh reveal (P1). - De-export defaultRedact, SecretInput, RevealRenderProps (knip: unused exports); they remain internal. - Drop an `as HTMLInputElement` cast in the ApiKeysSection test in favor of the exact masked-string assertion (which already covers length).
…daction - Add tests asserting Copy and Input place the REAL value on the clipboard while the snippet is still masked (the core copy-without-reveal promise). - Add McpInstallPanel consumer tests: the access key is masked in both the command and JSON-fallback snippets, and revealed only on an explicit action. - Clamp defaultRedact's visible prefix to half the value so short inputs are still masked instead of shown in full. - Redact longest secrets first so overlapping real values are order-independent. - Re-export defaultRedact and use it in the ApiKeysSection test instead of a local re-implementation. - Document the intended residual: MCP deep links carry the (encoded) key in their href by necessity; only the JSON fallback is reveal-gated.
brandon-pereira
requested review from
a team,
bot-hyperdx and
jordan-simonovski
and removed request for
a team and
bot-hyperdx
August 24, 2026 22:16
jordan-simonovski
approved these changes
Aug 25, 2026
MikeShi42
added a commit
that referenced
this pull request
Aug 28, 2026
…ted context The reviewer had no access to this repo's conventions at all. Verified with a canary rather than assumed: with `--setting-sources user` the model receives neither CLAUDE.md nor AGENTS.md; without the flag it receives AGENTS.md via the `@AGENTS.md` import but still not agent_docs/code_style.md, because AGENTS.md links that file rather than @-importing it. So the conventions never reached the model by any path. Dropping the flag is not an option: the working tree is fork-authored and that flag is what stops a PR injecting instructions into a job holding ANTHROPIC_API_KEY and a write token. So the conventions are fed in from the base-pinned trusted checkout instead. - agent_docs/code_style.md is added to the trusted sparse-checkout. Explicitly: this checkout is non-cone (it was set that way for .nvmrc), so root and nested files do not come along for free. - `Materialize review context` appends a `## Repository conventions` section holding that file, deliberately OUTSIDE the untrusted fence and labelled authoritative, since it comes from the base branch rather than the PR. The prompt's trust boundary gains a matching carve-out, so the reviewer does not distrust it along with everything else in the checkout. - The prompt gains a `Project conventions` pass. Feeding a rules document nobody is asked to apply would have changed nothing. - trusted-hash.sh covers the conventions file, so editing the rules invalidates the gate. That does mean a rules change re-reviews every open PR, which is the right trade for a change in what "correct" means. AGENTS.md is deliberately NOT included. It is 19.5KB of process guidance -- dev setup, running tests, changelog generation, merge conflicts -- with very little a reviewer can act on, and adding it would roughly double the token cost for that. code_style.md is 17KB of directly reviewable rules, several marked REQUIRED, and maps onto misses measured on the eval set: an icon-only ActionIcon with no accessible label (#1609) and a hand-rolled control that should have used a Mantine variant (#2988) are both REQUIRED sections of it. Two caveats. This is unevaluated: the harness passes the same `--setting-sources user`, so the 40% figure was measured without conventions too -- this is a new variable, not a restored one, and prior rounds showed added context is not free. And a PR that edits the conventions is still reviewed against the base copy, which is correct for trust and slightly wrong for that PR. 35 tests. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
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
2 participants
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.
What & why
Credentials were rendered in plain text across the app — the Team Settings API-key cards showed the full key, and the MCP install snippets inlined the personal access key directly into the copyable command/JSON. That's a shoulder-surfing risk on a screen-share or in a screenshot.
This adds a shared
RevealSnippetcomponent that masks secrets in a snippet until the user explicitly reveals them, while copy always hands back the real value (so nobody has to reveal a key just to copy it). It's adopted across every credential surface touched here:Design
RevealSnippetis a compound component —RevealSnippet.Code,.Reveal,.Copy, and a self-contained.Inputshare reveal state via context, so the surrounding layout (header row, panel chrome, button placement) is entirely up to the caller.Masking is driven by a single
secretsprop, with a deliberately safe default:secrets[][value](bare string)defaultRedact(prefix + dots) — for a bare API key[[real, redacted], …]Bare-string and pair entries can be mixed, and
nullentries are skipped sosecrets={[maybeKey]}is safe (e.g. a ClickHouse Cloud deployment with no key). The masked stand-in is padded to the real length so the field width doesn't jump on reveal.Screenshots
Testing
RevealSnippet(masking, reveal toggle, bare-string vs. pair vs. null secrets,canReveal=false, render-prop override) and updatedApiKeysSectiontests (masked value on the input, inline reveal).tsc --noEmit, ESLint, and unit tests all pass.