3 Commits

Author	SHA1	Message	Date
Molecule AI Core-QA	25866ec200	fix(workspace/OFFSEC-003): correct boundary wrapping + add closer truncation ... Two bugs fixed in tool_delegate_task wrapping logic: 1. Wrapping used raw _A2A_BOUNDARY_START/_END markers, which appeared in the output alongside the escaped form of the peer content (e.g. "[A2A_RESULT_FROM_PEER]\n[/ A2A_RESULT...]"). Fixed: wrap with _A2A_BOUNDARY_START_ESCAPED/_END_ESCAPED so the output contains no raw closer that could confuse downstream parsers. 2. A malicious peer could inject a fake closer ([/A2A_RESULT_FROM_PEER]) to make legitimate content appear truncated. Fixed: truncate at the raw closer BEFORE sanitization (truncation loses the raw form, so escaping afterward cannot retroactively remove it). Also fixes 10 regressions in test_a2a_offsec003_sanitization.py: tests were written expecting ZWSP (U+200B) escaping but implementation uses "[/ " prefix. Updated test invariants to match actual behavior. Also fixed 5 tests using [A2A_ERROR] in summary fields (not a boundary marker — no escaping applied) and updated test assertions in test_a2a_tools_impl.py and test_delegation_sync_via_polling.py to expect escaped wrapper forms. Cherry-picked fix/test-stdio-function-name (e478b5b2) from main: renamed _warn_if_stdio_not_pipe → _assert_stdio_is_pipe_compatible and added deprecated alias, fixing dangling monkeypatch targets that caused 5 test failures (issue #957). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>	2026-05-14 21:01:34 +00:00
Molecule AI Infra-Runtime-BE	635a42745a	fix(workspace): OFFSEC-003 — separate sanitize vs. wrap, fix tool_delegate_task (#477) ... Co-authored-by: Molecule AI Infra-Runtime-BE <infra-runtime-be@agents.moleculesai.app> Co-committed-by: Molecule AI Infra-Runtime-BE <infra-runtime-be@agents.moleculesai.app>	2026-05-11 15:10:25 +00:00
Molecule AI Infra-SRE	a205099652	fix(security): OFFSEC-003 — boundary-marker escape + shared sanitizer ... Root cause (from infra-lead PR#7 review id=724): Sanitization in PR#7 wrapped peer text in [A2A_RESULT_FROM_PEER] markers, but the markers themselves were not escaped — a malicious peer could inject "[/A2A_RESULT_FROM_PEER]" to close the trust boundary early, making subsequent text appear inside the trusted zone. Fix: - Create workspace/_sanitize_a2a.py (leaf module, no circular import risk) with shared sanitize_a2a_result() + _escape_boundary_markers() - _escape_boundary_markers() escapes boundary open/close markers in the raw peer text before wrapping (primary security control) - Defense-in-depth: also escapes SYSTEM/OVERRIDE/INSTRUCTIONS/IGNORE ALL/YOU ARE NOW patterns (secondary, per PR#7 design intent) - Update a2a_tools_delegation.py: import from _sanitize_a2a; wrap tool_delegate_task return and tool_check_task_status response_preview - Add 15 tests covering boundary escape, injection patterns, integration shapes (workspace/tests/test_a2a_sanitization.py) Follow-up (non-blocking, noted in PR#7 infra-lead review): - Deduplicate if a2a_tools.py also wraps (currently handled in delegation module only — callers get sanitized output regardless) - tool_check_task_status: consider sanitizing 'summary' field too Closes: molecule-ai/molecule-ai-workspace-runtime#7 (wrong-repo PR that this supersedes) Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>	2026-05-11 02:16:09 +00:00