Repository navigation
chore: Make the Bugbot approval receipt producer trust boundary explicit and reproducible (#980) - #992
Conversation
|
To preview the documentation for this pull request, visit the following URL:
|
There was a problem hiding this comment.
Current status: All findings originating in this review are resolved.
Last reconciled on6efac73. See aggregate Bugbot status.
🤖 Bugbot review snapshot
Bugbot reported 1 potential problem when commit 79396c2 was analyzed. This snapshot is historical; use the status block above for current state. 1 finding is linked to changed code.
Findings
- low: The evidence-failure case can pass without publishing its assessment —
quality/bugbot/evidence/issue-980/packaged-trust.cjs:185
Bugbot: review complete
Pull request · Verified commit · Current review checks · Workflow run Implementation progress and pending work · |
…plicit-and-reproducible: assert assessment publication separately from blocked evidence (#980)
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## develop #992 +/- ##
===========================================
+ Coverage 95.25% 95.66% +0.41%
===========================================
Files 1172 1563 +391
Lines 39236 49137 +9901
Branches 11685 15352 +3667
===========================================
+ Hits 37373 47006 +9633
- Misses 698 794 +96
- Partials 1165 1337 +172 🚀 New features to boost your workflow:
|
Copilot approval · Human review neededStatus: This pull request is outside the configured approval scope. Technical evidenceRevision: |
The packaged approval consumer authenticates run/App/SHA/artifact/check context but does not independently attest the producer executable or the truth of a fabricated clean receipt. This audit makes that operator trust assumption explicit and preserves the refuted broad protected-path approval exploit claim.
Fourteen localhost-only scenarios exercise the actual unchanged Action bundle with every PR-head read current: protected bundle/workflow/action edits, foreign identities, prior policy/attempt, ordinary authorized data, same-SHA rollback and existing human decisions. Fake assessment/approval POSTs are counted explicitly; real GitHub writes and qualification model calls are zero. Source approval behavior, production mode and consumer templates are unchanged. Independent producer code attestation remains a prerequisite before any future proposed enforcement.
Validation: 14/14 packaged scenarios (final exact driver), 236 existing consumer/domain/use-case tests, all 141 shadow controls, full mandatory 943 suites / 18,695 tests plus build/lint, specification/catalog/docs validators and Graphify. Bundles are byte-unchanged; whole-base source diff check passes. Initial sandbox bind failure and every later attempt remain as exact ledgers/compressed raw logs.
The final audit independently asserts a published result and exactly one received fake assessment POST for every scenario. A causal control makes the blocked-evidence assessment POST return 403: the old assertion set falsely reports 14 passed; the corrected driver rejects the same input. Both ledgers and raw outputs remain. All 141 shadow controls and the full 943-suite / 18,695-test build/lint hook pass again. On final SHA
6efac73417df179a3153b526ae6df9b0b53c976b, native Commit, Review, Verification, Branch Sync and static CI succeed; the bot resolved its earlier publication-assertion finding. Build, coverage and aggregate CI remain pending. No default installation, native canary, profile qualification or activation is claimed.Closes #980. Refs #950, #951, #977.