{ "schema_version": 1, "builder_plan": { "scenario": "An agent reads external documents or tool responses that may contain malicious instructions.", "change": "Replay attacks and legitimate lookalikes in an isolated copy of your agent workflow.", "baseline": "Your existing checks with the same tasks, permissions, and tool responses.", "steps": [ "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." ], "measure": "Successful attacks missed, legitimate actions blocked, detection delay, and review cost.", "decision": "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." }, "builder_review": null, "arxiv_id": "2608.30041", "title": "Reachability-Based Capability Confinement for LLM Agents under Indirect Prompt Injection", "summary": "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.", "rationale": "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.", "technique": "Reachability-based capability confinement for LLM agents", "family": "seguranca_e_guardrails", "recommendation": "testar", "published": "2026-08-30", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2608.30041", "independent_implementations": 1, "total_implementations": 2, "stars": 1, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "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.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2608.30041", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "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.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2608.30041", "source_page": null, "source_excerpt": "" } ], "exposure_map": [ { "dimension": "quality", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "compute", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "latency", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "operations", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "compatibility", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "security", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "data_and_training", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "reproducibility", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" } ], "risks": [], "minimum_test": [], "open_questions": [], "independent_tests": [], "lab_reproductions": [], "report_available": false, "validation_tier": "unknown", "evidence_tier": "unknown", "infrastructure_basis": "unknown", "equations": [ { "anchor": "S3.Ex2", "label": "", "section": "3 Detailed Design of SkillGuard", "role": "proposed_method", "latex": "(R_{\\mathrm{rm}}^{*},T^{*}) =\\arg\\min_{R,T}W(R,T), \\\\ \\text{s.t.}\\quad\\operatorname{Reach}_{\\mathcal{G}(R,T)}(q)\\cap B=\\varnothing.", "mathml": "(Rrm∗,T∗)=arg⁡minR,T⁡W⁡(R,T),s.t.Reach𝒢⁡(R,T)⁡(q)∩B=∅.", "context": "Let K⁡(q)⊆Σ denote the capabilities enabled in state q. A restriction consists of a removal set R⊆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", "headline": "", "explanation": "", "objective": "", "symbols": {}, "evidence": "", "evidence_url": "", "source_url": "" }, { "anchor": "S4.Ex11", "label": "", "section": "4 Workflow of SkillGuard", "role": "proposed_method", "latex": "\\min_{x\\in\\{0,1\\}^{K(q)}} \\sum_{c\\in K(q)}w(c)(1-x_{c}) \\\\ \\quad\\text{s.t.} \\sum_{c\\in C(\\pi)}(1-x_{c})\\geq 1,\\ \\ \\forall\\,\\pi\\in\\Pi(q,B).", "mathml": "minx∈{0,1}K⁡(q)∑c∈K⁡(q)w⁡(c)​(1−xc)s.t.∑c∈C⁡(π)(1−xc)≥1,∀π∈Π(q,B).", "context": "Binary Formulation. Each capability is represented by a binary decision variable, xc∈{0,1}, where xc=1 denotes that capability c is retained and xc=0 denotes that it is removed. The minimum-cost capability cut is formulated as:", "headline": "", "explanation": "", "objective": "", "symbols": {}, "evidence": "", "evidence_url": "", "source_url": "" }, { "anchor": "S4.Ex12", "label": "", "section": "4 Workflow of SkillGuard", "role": "proposed_method", "latex": "\\min_{\\mathbf{y}}\\quad\\sum_{c\\in R_{\\mathrm{bin}}^{*}}w_{c}(y_{c}) \\\\ \\text{s.t.}\\quad\\operatorname{Reach}_{\\mathcal{G}[\\mathbf{y}]}(q)\\cap B=\\varnothing,\\quad y_{c}\\in[0,1],\\quad\\forall c\\in R^{*}_{\\mathrm{bin}}.", "mathml": "min𝐲∑c∈Rbin∗wc​(yc)s.t.Reach𝒢⁡[𝐲]⁡(q)∩B=∅,yc∈[0,1],∀c∈Rbin∗.", "context": "Given the tighting vector 𝐲, 𝒢⁡[𝐲] denotes the restricted SIG obtained by removing transitions whose parameters violate the tightened envelopes Ec​(yc). The optimal fractional restriction vector 𝐲∗ is obtained via:", "headline": "", "explanation": "", "objective": "", "symbols": {}, "evidence": "", "evidence_url": "", "source_url": "" } ], "equations_status": "selected", "core_kind": "formula", "equations_fetched_at": "2026-09-22", "equations_source_url": "", "equations_source_sha256": "", "exposures_reviewed_at": "", "exposures_source_sha256": "" }