carlos / platform Public

spec: record Astra round 5, accept

Summary

Round 5 found nothing at P1 or P2 and accepted rev 5. Findings committed
unedited. §16 remains Paul's decision.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

claude 2 weeks ago bd74c94e0997c371ff147583e0ac79aea802adb7 parent e764b20

2 files changed, +13 −0

docs/superpowers/specs/2026-10-05-deploy-credential-design.md +3 −0
docs/superpowers/specs/CODEX-FINDINGS-DEPLOY-CREDENTIAL-R5.md +10 −0

CI passed — run details

diff --git a/docs/superpowers/specs/2026-10-05-deploy-credential-design.md b/docs/superpowers/specs/2026-10-05-deploy-credential-design.md

index 2cdee8de3..edc4fd18d 100644

--- a/docs/superpowers/specs/2026-10-05-deploy-credential-design.md

+++ b/docs/superpowers/specs/2026-10-05-deploy-credential-design.md

@@ -41,6 +41,9 @@

of the directory store than it provides, and §12's randomized test

asserted something G1 and G2 do not imply. Rev 5 qualifies G2 by backend

and corrects the test's oracle. Changes are marked "(codex R4-N)".

+- Rev 5 (e764b20c) went back. Astra's round 5

+ (`CODEX-FINDINGS-DEPLOY-CREDENTIAL-R5.md`) found nothing at P1 or P2 and

+ accepted. Rev 5 is unchanged by it. §16 is still open for Paul.

**Discussion:** carlos/platform#229.

**Code read at:** main `8d8e4bae`. Every line reference below is at that commit.

diff --git a/docs/superpowers/specs/CODEX-FINDINGS-DEPLOY-CREDENTIAL-R5.md b/docs/superpowers/specs/CODEX-FINDINGS-DEPLOY-CREDENTIAL-R5.md

new file mode 100644

index 000000000..f2baffbb1

--- /dev/null

+++ b/docs/superpowers/specs/CODEX-FINDINGS-DEPLOY-CREDENTIAL-R5.md

@@ -0,0 +1,10 @@

+Reviewed `e764b20c`, Round 5. No files modified; no tests run.

+**No P1 or P2 findings.**

+- **R4-1 is closed by qualification.** §7.2 now limits directory-store G2 to writers sharing one handle, matching the mutex and comparison/rename implementation. §14 explicitly separates the cross-handle defect. The S3 condition matches [AWS’s documented conditional-write semantics](https://docs.aws.amazon.com/AmazonS3/latest/userguide/conditional-writes.html).

+- **R4-2 is closed.** §12’s oracle checks the pipeline actually returned by the authorization read. Installing a pipeline before the pointer read, then removing it before authorization, no longer produces a false failure.

+- **Both mutants violate the revised oracle.** Moving authoritative authorization before the pointer read violates G1. Making the write unconditional violates G2 on a schedule where another writer changes the pointer between its read and replacement. Printed, replayable failing schedules are explicitly required.

+- I found no new regression or incorrect citation in the revision. §16 remains Paul’s decision; this review does not resolve it.

+**Verdict: accept.**

\ No newline at end of file