Agent Approval Laundering: Transitive Effects Beyond the Approved Invocation

An agent reads external documents or tool responses that may contain malicious instructions.

What you would change

Replay attacks and legitimate lookalikes in an isolated copy of your agent workflow.

Replay attacks and legitimate lookalikes in an isolated copy of your agent workflow.

Compare against

Your existing checks with the same tasks, permissions, and tool responses.

1. Locate the paper's implementation and check its license, model support, data needs, and hardware requirements. Stop here if you cannot obtain the required artifacts.
2. Save a small set of representative product tasks and expected outcomes. Keep evaluation tasks out of any training or tuning, and choose quality and cost limits before running.
3. Run the baseline and the proposed change on identical inputs. Keep other settings fixed and repeat paired runs to expose variation.
4. Save per-task outputs, failures, timings, and all usage costs. Review regressions before expanding the test.

What to measure

Successful attacks missed, legitimate actions blocked, detection delay, and review cost.

When to keep it

Keep the change only if it meets your preselected quality and latency limits and improves the metric you care about after setup and operating costs. Otherwise retain the baseline. This is an exploration plan, not a validated recommendation.

recommendation

test

research area

safety and guardrails

technical change

Effect-bound approval records for agent tool calls

published

2026-09-23

The reader likely operates coding agents and MCP tools in production, and the Claude Code PreToolUse integration makes this deployable on constrained infrastructure, but the small evaluation base and nonzero residual rate justify a scoped pilot rather than full adoption.

1. claim 01Paperraft inference
   Read the brief interpretation

   It replaces invocation-level approval logging, where the human decision is bound only to the entry command or MCP call, with closure-bound records that commit a frozen, source-backed prediction of the workflow's transitive effects (lifecycle hooks, file writes, network access) before authorization. The cost is added tooling in the permission path (a PreToolUse hook that computes and carries the frozen record), pre-execution analysis latency, and the engineering burden of maintaining effect predictions and provenance for each tool. Prediction errors are the residual failure mode: the holdout still left 3 of 10 residual effects unbound, so effects outside the six modeled classes or missed by the predictor remain invisible to the approval record, and the evidence base is small (111 pairs, 17 holdout workflows, one product integration).

   This is a Paperraft inference. It is not pinned to a verified PDF passage.

   Inspect the source
2. claim 02Paperraft inference

   On 17 prespecified holdout workflows, frozen source-backed effect predictions achieve 0.926 macro recall and 0.941 macro precision, and binding them to approvals cuts residual (unrecorded) effects from 10 to 3; on 111 fixed approval-object/trace pairs, decision-time metadata reduces residual records from 40 to 13.

   This is a Paperraft inference. It is not pinned to a verified PDF passage.

   Inspect the source