Run repository tests on the host by default; make sandboxed test execution explicit opt-in (#535) - #540
Merged
Conversation
…test execution explicit opt-in (#535) - runtime-validation.md: new "Repository test execution backend" section (classification, host default, explicit sandbox, no fallback, not executed != failed); pointers from trust model, Safety gate, targeted mode - trusted-host-execution.md: separate `run_repository_tests_in_sandbox` request (structured + closed NL vocabulary, denial phrases count, conflicts resolve to sandbox); `host` provenance - sandbox runner: launcher exec failure (seatbelt 71/execvp, bwrap execvp, docker 126/127) maps to `unavailable`, never `failed` - reference model + scenario tests parameterized over both Skills; NL corpus sandbox-request case set; telemetry `host` provenance - runbooks 8a/active/passive, skill.yaml, capability.yaml, review-summary, feature docs; threat catalog SBOX-014 (accepted risk) and SBOX-015 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…d the sandbox-request question guard (#535) - runtime-validation.md "Purpose and boundary", local-review.md working-tree paragraph, and github repository-checkout.md now except admitted repository test commands and point to "Repository test execution backend" - reference model: the question guard matches one sentence only, so an earlier sentence can no longer drop a later explicit sandbox request - wiring tests pin the carve-outs; #223 local runbook hash re-captured Co-Authored-By: Claude Opus 5.5 <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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fixes #535
What
host, in both Skills. An explicit per-invocation sandbox request (run_repository_tests_in_sandboxor a small closed phrase set, incl. existing denial phrases likesandbox only) makes it sandbox-only with no host fallback. A sandbox that cannot launch the test toolchain isunavailable, neverfailed. Lint/build/type-check, unclassifiable commands, and generated reproductions keep the sandbox-required contract unchanged. Canonical:runtime-validation.md"Repository test execution backend",trusted-host-execution.md"Repository test sandbox request".scripts/sandbox/runner.pymaps a launcher exec failure (seatbelt 71/execvp(), bwrapexecvp, docker 126/127) tounavailable; a started non-zero exit staysfailed.github-pr-reviewof an untrusted fork runs its tests on the host unless the sandbox is requested.review-summary.md, both runbooks (step 8a / active / passive),repository-checkout.md, andskill.yamlcomments.CHANGELOG.md(seedocs/RELEASE.md)unavailableinstead offailed.Validation
tests/unit3117 OK,tests/policy1898 OK (re-run after mergingmain),tests/repository8 OK,tests/integration138 OK; threat-model and traceability validators, Skill metadata validation, packaging, and package-manifest--checkpass.sandbox-exec), a separate NL sandbox-request corpus set, and runbook wiring tests.Review
local-code-reviewinitial review ona52b7a1: CHANGES REQUIRED (P1: unchanged "dormant validation" statements contradicted the host default; P2: sandbox-request question guard spanned sentences). Both fixed in5fd3f62.local-code-reviewre-review on5fd3f62: REVIEW CLEAN (P0 0, P1 0, P2 0).🤖 Generated with Claude Code