Repository navigation
The latest security review agent in ship is triggering usage policy API denials from Anthropic #1899
Description
The Usage Policy block happens because the repo’s security regression tests use deliberate attack payloads (rm -rf ~, command substitution, injection strings). When /ship runs adversarial review, red team, or the security specialist subagent, Opus tries to “think like an attacker” against that material and gets blocked by real-time cyber safeguards.
There needs to be a way to segregate those kinds of tests so that they are manually run and approved
Activity
garrytan commentedon Jun 8, 2026
Fixed in #1911 (shipping in v1.57.6.0). The Claude adversarial subagent now carries an authorized-defensive-testing preamble and reads test/fixture files in summary mode, so it no longer trips Anthropic's usage policy when reviewing the repo's own attack-payload fixtures. Fixed at the prompt source (
scripts/resolvers/review.ts). Thanks for the report. Will auto-close when the PR merges.v1.57.6.0 fix wave: 8 community bugs (4 security guards failing open) (…
fix(review): extend garrytan#1899 defensive framing to red-team + sec…