You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: extensions/business/cybersec/red_mesh/AGENTS.md
+1-1Lines changed: 1 addition & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -302,7 +302,7 @@ Only append entries for critical or fundamental RedMesh backend changes, discove
302
302
303
303
### 2026-03-16T20:40:00Z
304
304
305
-
- Change: introduced a dedicated LLM payload-shaping boundary in [`mixins/llm_agent.py`](./mixins/llm_agent.py) so RedMesh no longer sends the full aggregated report directly to the LLM path.
305
+
- Change: introduced a dedicated LLM payload-shaping boundary in [`mixins/llm_agent_mixin.py`](mixins/llm_agent_mixin.py) so RedMesh no longer sends the full aggregated report directly to the LLM path.
306
306
- Change: added network and webapp-specific compact payload shaping, finding deduplication/ranking/capping, analysis-type budgets, and runtime payload-size observability.
307
307
- Verification: the known failing job `a3a357bc` dropped from `303,760` raw bytes to `21,559` shaped bytes for `security_assessment` and completed manually in `38.97s` on rm1 instead of timing out.
308
308
- Horizontal insight: RedMesh archive/report data and LLM reasoning data must remain separate contracts; future LLM work should extend the bounded payload model rather than re-coupling the agent to raw archived aggregates.
0 commit comments