claude-code - 💡(How to fix) Fix Agent tool: parallel agents with destructive operations cause data loss — needs sequencing guardrails [1 comments, 2 participants]

The Agent tool allows spawning parallel agents with no dependency awareness. When Agent A (transfer) and Agent B (cleanup/delete) operate on the same source files, there is no mechanism to enforce ordering. The model should have sequenced them, but the tool doesn't warn or prevent this pattern.

1. When multiple agents are spawned in the same message and one performs destructive operations (rm, delete, move) on paths that another agent reads from, the system should warn the model or enforce sequential execution.

TL;DR

To prevent data loss, ensure that agents performing destructive operations on a path do not run in parallel with agents reading from the same path by implementing a dependency mechanism or sequential execution.

Guidance

• Identify and document all instances where agents may be spawned in parallel with potential read-write conflicts, and manually sequence them if necessary.
• Modify the Agent tool to either warn about or prevent parallel execution of agents with destructive operations on paths that other agents are reading from.
• Consider adding a depends_on parameter to allow for explicit sequencing of dependent agents while preserving parallelism for independent tasks.
• Review and update the Agent tool documentation to explicitly caution against running destructive operations in parallel with read operations on the same paths.

Example

A potential code change could involve introducing a depends_on parameter, for example:

agent_a = Agent("read_files", path="X")
agent_b = Agent("delete_files", path="X", depends_on=[agent_a.id])

This would ensure that agent_b only starts after agent_a has completed.

Notes

The solution requires careful analysis of existing agent workflows to identify potential conflicts and may need significant updates to the Agent tool and its documentation. The introduction of a dependency mechanism can add complexity but is necessary to prevent data loss.

Recommendation

Apply a workaround by manually sequencing agents with potential read-write conflicts until a permanent fix with dependency awareness is implemented, as this will immediately prevent further data loss.