Repository navigation
fix(redaction): keep Infisical secret names visible, and correct the NO_PROXY docs - #9867
waleedlatif1 wants to merge 2 commits into
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
Folded into the single round-2 release cleanup PR (fix/release-final-leftovers-1110), which takes both commits here unchanged: the reviewed secretKey redaction fix and the networking.mdx NO_PROXY wording. Closing so there is one PR. |
[Critical impact] The PR changes a shared redaction rule used across several ways people inspect run data. Restore masking for Findings
SummaryThis PR keeps Infisical secret names visible and corrects the outbound proxy documentation.
Reviews (1) · Last reviewed commit: "docs(networking): the SSRF guard matches..." · Reviewed by Greptile |
| // Convex's `deployKey`. A bare `secretKey` is left out: Infisical returns each | ||
| // secret's name in that field, and redacting it would hide every listed name. | ||
| /^.*deploy[_-]?key$/i, |
There was a problem hiding this comment.
Removing the whole secretKey suffix pattern also stops hiding bedrockSecretKey, which holds an AWS secret access key. When a user types that key directly into a Bedrock block, the resolver keeps the literal value and the block logger passes it through redactApiKeys. The key now appears in logged inputs and the console.
Keep bare secretKey visible for Infisical, but retain coverage for credential fields such as bedrockSecretKey and restore its regression test.
How this was verified: The Bedrock password field flows into logged block inputs, while the placeholder projection preserves directly typed values and the remaining key patterns do not match bedrockSecretKey.
Summary
secretKeypattern after review (Infisical returns each secret's name insecretKey), but that commit never reached the PR before it was squash-merged. Staging therefore redacts every Infisical secret name in persisted logs and console output. This restores the reviewed version: only Convex'sdeployKeyis added, andsecretKeyis back on the non-sensitive list with the reason in the commentfetch, which fix(release): keyless subject uniqueness, NO_PROXY IPv6/CIDR, proxied boot secret, and remaining audit follow-ups #9866's curl-styleNO_PROXYmatching made untrue. It now says the guard matches like curl and links to the section that explains bothType of Change
Testing
bunx vitest run lib/core/security/redaction.test.ts— 76 passed; restoring thesecretKeypattern turns thesecretKeynon-sensitive case redcheck:docs,check:site-urlslint:check+type-checkvia the pre-push hook; full gate left to CIChecklist
test-auditauthoring gate)