The Over-Tooled Agent

Part I

Problem Definition

An agent can expose too many tools, overly broad tool functions, or credentials whose effective permission closure exceeds the current task [R1] [R2].

This note treats the over-tooled agent as a bounded control question. It does not infer universal product behavior from a paper, standard, interface screenshot, or single test. A demonstrated result applies only to the cited configuration; an authoritative source defines a requirement or design direction but does not certify an implementation. ACE therefore asks whether the tested system can produce the required behavior and evidence under declared versions, policies, topology, identities, and failure conditions.

The operational distinction is between a claim and a control that can be re-performed. The positive control is: A summarization agent receives one read-only document tool and no external egress capability. The adversarial condition is: The same agent receives file, shell, browser, email, and admin tools and combines a document injection with outbound transfer. A useful result must show both that authorized work remains possible and that the prohibited path is stopped before an irreversible side effect. Missing fields are classified as insufficient evidence, not silently converted into a pass.

Real-world impact

• A single injection can compose individually benign tools into an exfiltration or mutation path.
• Frequent low-value approvals can create fatigue without reducing the effective privilege closure.
• A control claim without versioned evidence can mislead procurement, audit, incident response, and system owners.
• Failing closed without a positive control can conceal a denial-of-service design rather than demonstrate trustworthy behavior.

Part II

Mitigation Direction

Research-backed direction

1. 01
   research proposed

   Expose only the tools required for the current workflow step and remove open-ended shell, URL, or SQL capabilities by default [R1].

2. 02
   research proposed

   Audit actual file, network, process, and data capabilities rather than trusting tool names or descriptions [R3].

3. 03
   research proposed

   Apply chain-level policy to combinations such as read plus external send, not only to isolated calls [R2].

LogionOS engineering mapping

Implementation hypotheses only. No production validation is claimed.

1. 01
   implementation hypothesis

   Add a versioned policy object for the over-tooled agent and keep its decision inputs outside model-writable context.

2. 02
   implementation hypothesis

   Generate a signed technical receipt linking actor, request, policy version, decision, enforcement point, and observed outcome.

3. 03
   implementation hypothesis

   Export missing evidence explicitly as insufficient evidence and limit every claim to the tested configuration.

Part III

Consequences and Research Agenda

Consequences

• A single injection can compose individually benign tools into an exfiltration or mutation path.
• Frequent low-value approvals can create fatigue without reducing the effective privilege closure.
• A failed acceptance test requires the related capability claim to remain not demonstrated or insufficient evidence.
• A passing test supports only the declared configuration and does not establish universal safety.

Second-order effects

• Stronger enforcement can increase latency, state, operational dependencies, and legitimate denials.
• More evidence can increase privacy and retention exposure unless raw content is minimized and access-controlled.
• A detector or policy service can become a new failure point and must have explicit fail behavior.
• Attackers may adapt to published checks, so the public test direction should be paired with private regression variants.

Limitations

• Several cited AI-agent and reasoning-security sources are preprints or bounded experiments; they are identified as such in the references.
• The proposed ACE acceptance test has not yet been run across all incumbent and AI-native implementations.
• Cryptographic integrity proves that an artifact was not altered after commitment; it does not prove that the artifact was true, complete, or correctly interpreted.
• Legal and contractual applicability remains deployment- and jurisdiction-specific.

Open research questions

1. How should combined tool capability be measured before execution?
2. Can capability budgets adapt without silently expanding during a long workflow?
3. Which evidence fields are mandatory for a demonstrated result, and which may be not applicable?
4. How should continuous regression detect policy, model, tool, and provider drift after the initial test?

Sources

References

1. [R1]
   OWASP LLM06: Excessive Agency

   published · authoritative-standard

2. [R2]
   MCP Security Best Practices

   published · authoritative-standard

3. [R3]
   Auditing MCP Servers for Over-Privileged Tool Capabilities

   preprint · original-paper

Record

Publication Record

Recommended citation

Ma, Chris. “The Over-Tooled Agent.” ACE Research Note ACE-RN-2026-007, v1.0, 2026.

Corrections

No corrections recorded.

Organizational disclosure

ACE Research and LogionOS share organizational affiliation. LogionOS mappings in this note are implementation hypotheses, not independently validated product claims.

Evidence boundary

This note synthesizes cited public research and defines an ACE acceptance direction. It does not report a completed cross-vendor experiment unless explicitly stated, and it contains no private ACE prompts, holdout identifiers, customer data, or raw model responses.