fix(agent): seal document approval content

This commit is contained in:
RaresKeY
2026-08-15 07:01:36 +00:00
parent 2b72531eaa
commit 7a138e8a3f
7 changed files with 151 additions and 10 deletions
+69
View File
@@ -1000,6 +1000,75 @@ def test_tainted_document_edit_without_active_target_cannot_be_approved(monkeypa
assert "ask_user" not in blocked[0]
def test_tainted_document_approval_seals_current_content(monkeypatch):
from types import SimpleNamespace
from src.prompt_security import untrusted_context_message
from src.tool_approvals import document_content_digest
import src.agent_loop as agent_loop
monkeypatch.setattr(
agent_loop,
"get_setting",
lambda key, default=None: default,
raising=False,
)
monkeypatch.setattr(agent_loop, "get_mcp_manager", lambda: None, raising=False)
monkeypatch.setattr(agent_loop, "estimate_tokens", lambda *args, **kwargs: 10)
async def fake_stream(*args, **kwargs):
yield "data: " + json.dumps({
"delta": "```update_document\nreplacement\n```",
}) + "\n\n"
yield "data: [DONE]\n\n"
async def should_not_execute(*args, **kwargs):
raise AssertionError("unapproved document edit reached executor")
monkeypatch.setattr(agent_loop, "stream_llm_with_fallback", fake_stream)
monkeypatch.setattr(agent_loop, "execute_tool_block", should_not_execute)
active_document = SimpleNamespace(
id="document-7",
title="Draft",
language="markdown",
current_content="original",
version_count=4,
)
events = _collect_agent_events(
agent_loop.stream_agent_loop(
"http://local.test/v1",
"small-local-model",
[
{"role": "user", "content": "update this document"},
untrusted_context_message("stored context", "untrusted"),
],
active_document=active_document,
session_id="document-approval-session",
owner="alice",
max_rounds=1,
relevant_tools={"update_document"},
)
)
approval = next(
event["ask_user"]
for event in events
if event.get("ask_user", {}).get("kind") == "tool_approval"
)
pending = agent_loop.tool_approval_store.peek(approval["approval_id"])
assert pending is not None
assert pending.document_id == "document-7"
assert pending.document_version == 4
assert pending.document_digest == document_content_digest("original")
agent_loop.tool_approval_store.consume(
pending.approval_id,
decision="deny",
owner="alice",
session_id="document-approval-session",
)
def test_approval_pause_does_not_trigger_teacher_takeover(monkeypatch):
from src.prompt_security import untrusted_context_message