fix(tool_runner): avoid leaking tool exception details (credit @sebastiondev #814)#823
Open
Bartok9 wants to merge 1 commit into
Open
fix(tool_runner): avoid leaking tool exception details (credit @sebastiondev #814)#823Bartok9 wants to merge 1 commit into
Bartok9 wants to merge 1 commit into
Conversation
Salvage of RobotecAI#814 by @sebastiondev. Exception messages raised by tool implementations were embeddedverbatim into ToolMessage.content returned to the agent. Log full details via logger.exception server-side and return only the tool name plus exception class name to the LLM/user. Adds offline regression tests proving secrets and hosts are not presented in ToolMessage content. Signed-off-by: Bartok9 <danielrpike9@gmail.com>
Bartok9
changed the title
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
ToolMessagereturned to the LLM/user).logger.exception.ToolMessage.content.Credit
Original investigation, root-cause analysis, and fix approach by @sebastiondev in #814. This salvage rebases the intent onto current
mainand adds tests so the property stays locked.Motivation
Tool failures from DB/HTTP/SSH libraries often embed hosts, paths, and credentials in
str(e). Feeding that intoToolMessage.contentsurfaces internals to the LLM prompt and any conversation UI.Verification
Notes for reviewers
Happy to close this in favor of #814 if preferred — the tests can be ported either direction.
Agent-Owner:
sera· Platform: hermes · Claim-TTL: 24h