Use serving assessments for adaptive Guardian review (#58075)

## What changed

- Match final serving assessments to their response and output item, then retain them with tool invocations and code cells across waits.
- Use matching assessments instead of asynchronous scoring for adaptive review: low-risk actions can skip synchronous review, while elevated-risk actions require it.
- Revalidate captured authorization and policy before approval. Require fresh review for stale or incomplete authorization, preserve explicit review requirements, and discard assessments when hooks rewrite tool input.
- Keep asynchronous scoring as the fallback when an assessment is missing or does not match.

## Testing

Add integration coverage for direct and nested tool calls, hook rewrites, explicit escalation, interrupted responses, catalog-driven policy, incomplete authorization, and assessment retention across code-cell yields and waits.

GitOrigin-RevId: 4346a227204137f018c03c34b302d6f1e9592c96
This commit is contained in:
jif
2026-10-29 12:54:39 +00:00
committed by copyberry
parent 6b3b3349f3
commit 9def09afa7
26 changed files with 592 additions and 97 deletions
+1 -1
View File
@@ -49,7 +49,7 @@ struct ToolCallTimingGuard {
pub(crate) struct ToolCallRuntime {
session: Arc<Session>,
// Tool calls may run later, so retain the step whose tool list advertised them.
step_context: Arc<StepContext>,
pub(crate) step_context: Arc<StepContext>,
tracker: SharedTurnDiffTracker,
parallel_execution: Arc<RwLock<()>>,
timing: Option<DirectToolTiming>,