Reachability-Based Capability Confinement for LLM Agents under Indirect Prompt Injection · Research brief

Request deep report Read original paper ↗View page data (JSON)Download research skill ↓Discuss this paper

For solo builders

A possible application for this research area. Paper-specific feasibility still needs a full-paper review.

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.

Try the idea on your product

A proposed first test. Set your budget and success criteria before running it.

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.

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.
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.
Run the baseline and the proposed change on identical inputs. Keep other settings fixed and repeat paired runs to expose variation.
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.

Download test plan ↓

abstract indexedPaperraft next step

Do not allocate compute yet.

This page uses the abstract and public adoption signal only. Request the deep report before relying on its recommendation.

Abstract-level screening only. Risks and infrastructure may not have been evaluated. This recommendation is not a reproduced result.

minimum test setup: not reported
published evidence setup: not reported
source-linked claims: 0 of 2
exposure checks: 0 of 8

Why it was shortlisted

The technical change, research area, and reason it entered the archive.

recommendation: test
research area: safety and guardrails
technical change: Reachability-based capability confinement for LLM agents
published: 2026-08-30

The defense adds zero inference cost and works with API-based backends the reader already uses, but its security guarantees depend on hand-written per-skill summaries and policies whose quality must be validated on the reader's own tool set before production adoption.

Central equations

Original equations, typeset for reading. Paperraft explanations connect the mechanism, its aim, and the evidence.

Let $K (q) \subseteq Σ$ denote the capabilities enabled in state $q$ . A restriction consists of a removal set $R \subseteq K (q)$ and a tightening map $T$ over the retained capabilities. Applying $(R, T)$ removes transitions associated with capabilities in $R$ and restricts transitions that violate the tightened steerability signatures specified by $T$ . Let $𝒢 (R, T)$ denote the resulting restricted SIG and let $W (R, T)$ denote the cost of applying the restriction. SkillGuard selects the minimum-cost restriction

(R_{rm}^{*}, T^{*}) = \arg \min_{R, T} W (R, T),

s.t. {Reach}_{𝒢 (R, T)} (q) \cap B = \emptyset .

unnumbered equation · §3 Detailed Design of SkillGuard · proposed method

Binary Formulation. Each capability is represented by a binary decision variable, $x_{c} \in {0, 1}$ , where $x_{c} = 1$ denotes that capability $c$ is retained and $x_{c} = 0$ denotes that it is removed. The minimum-cost capability cut is formulated as:

\min_{x \in {0, 1}^{K (q)}} \sum_{c \in K (q)} w (c) ​ (1 - x_{c})

s.t. \sum_{c \in C (π)} (1 - x_{c}) \geq 1, \forall π \in Π (q, B) .

unnumbered equation · §4 Workflow of SkillGuard · proposed method

Given the tighting vector $𝐲$ , $𝒢 [𝐲]$ denotes the restricted SIG obtained by removing transitions whose parameters violate the tightened envelopes $E_{c} (y_{c})$ . The optimal fractional restriction vector $𝐲^{*}$ is obtained via:

\min_{𝐲} \sum_{c \in R_{bin}^{*}} w_{c} ​ (y_{c})

s.t. {Reach}_{𝒢 [𝐲]} (q) \cap B = \emptyset, y_{c} \in [0, 1], \forall c \in R_{bin}^{*} .

unnumbered equation · §4 Workflow of SkillGuard · proposed method

equations from arXiv HTML · fetched 2026-09-22

Claims and evidence

A claim is source-linked only when the PDF page and matching excerpt are available.

claim 01Paperraft inference
Read the brief interpretation
SkillGuard replaces LLM-inference-based prompt-injection filters and content classifiers with a harness-level reference monitor that treats untrusted skill output as contamination and restricts the agent's future capabilities via a Skill Impact Graph and steerability signatures. It costs no additional model calls or token overhead, but requires authoring sound skill summaries, steerability signatures, and deployer-defined forbidden-state policies, plus integration into the agent harness. It can fail if skill summaries are incomplete or unsound, if policies misclassify which states are forbidden, or if fractional restriction parameters are tuned to preserve utility at the expense of residual attack paths, as the non-zero Slack attack success rates show.
This is a Paperraft inference. It is not pinned to a verified PDF passage.
Inspect the source
claim 02Paperraft inference
Under AgentDojo Tool Knowledge attacks, SkillGuard eliminates attack success on three of four suites for both backends and reduces it to 4.8% (Gemini) and 14.3% (Llama) on Slack, with no added model calls or token overhead.
This is a Paperraft inference. It is not pinned to a verified PDF passage.
Inspect the source

Exposure map

Author findings, Paperraft interpretations, and the next check. Missing analysis never counts as evidence of safety.

quality

Still unknown

Which tasks improve, and which regress against the baseline?

compute and memory

Still unknown

What memory, hardware, and total compute does the complete method require?

latency

Still unknown

Are startup time, typical response time, and slow requests measured separately?

operations

Still unknown

What must be monitored, maintained, and rolled back?

software compatibility

Still unknown

Does the implementation support your model, runtime, and hardware?

security

Still unknown

What new data access, permissions, or trust boundaries does it introduce?

data and training

Still unknown

What data preparation or training is required, and is it included in the cost?

reproducibility

Still unknown

Are code, versions, inputs, and evaluation scripts available?

Continue the investigation

A reading guide for safety and guardrails.

Editorial prompts, not findings from this paper or a validated test plan. Detailed analysis is still pending for the topics below.

What to inspect

Which assumptions does the proposed method need, and where might your workload violate them?

Plan a fair comparison

Identify the baseline, metric, dataset, and experimental settings in the paper. Which can you match?

Questions for adoption

What implementation and evaluation artifacts are available, and what would you need to supply?

Read the paper with these questions

Inside the paper

Preview the first 3 pages of the original PDF. Click the stack to turn the page.

The downloadable skill includes SKILL.md and evidence.json, with recorded findings and unanswered checks. TRY-IT.md adds a product test plan and a blank results sheet.

A reading aid for your agent. Findings still need validation.

Page 1 of 3

Open this page in the PDF ↗

Provisional research brief updated 2026-10-18. Paperraft has not reproduced this experiment.

For solo builders

What you would change

Try the idea on your product

Compare against

What to measure

When to keep it

Do not allocate compute yet.

Why it was shortlisted

Central equations

Claims and evidence

Under AgentDojo Tool Knowledge attacks, SkillGuard eliminates attack success on three of four suites for both backends and reduces it to 4.8% (Gemini) and 14.3% (Llama) on Slack, with no added model calls or token overhead.

Exposure map

quality

compute and memory

latency

operations

software compatibility

security

data and training

reproducibility

Continue the investigation

What to inspect

Plan a fair comparison

Questions for adoption

Inside the paper

Discuss this paper