When Tool Metadata Becomes an Instruction

Part I

Problem Definition

Tool descriptions, annotations, schemas, and returned metadata can contain attacker-controlled text that the model interprets as instructions rather than untrusted data [R1] [R2].

This note treats when tool metadata becomes an instruction 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: An approved tool manifest describes a read operation and returns structured records. The adversarial condition is: A tool description or hidden metadata asks the agent to ignore policy, retrieve secrets, and call an outbound tool. 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 trusted-looking tool catalog can become a supply-chain instruction channel.
• Description changes may alter agent behavior without changing executable code or credentials.
• 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

   Treat server-supplied annotations as untrusted display data and never as authorization policy [R1].

2. 02
   research proposed

   Pin approved tool manifests and require review when descriptions, schemas, or capabilities change [R2].

3. 03
   research proposed

   Use structured outputs and independent policy enforcement so injected text cannot authorize a side effect [R2] [R3].

LogionOS engineering mapping

Implementation hypotheses only. No production validation is claimed.

1. 01
   implementation hypothesis

   Add a versioned policy object for when tool metadata becomes an instruction 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 trusted-looking tool catalog can become a supply-chain instruction channel.
• Description changes may alter agent behavior without changing executable code or credentials.
• 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. Which metadata fields must be excluded from model context entirely?
2. How should enterprises review dynamic tool-list changes without freezing legitimate updates?
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]
   MCP Tools Specification

   published · authoritative-standard

2. [R2]
   OWASP MCP Tool Poisoning

   published · authoritative-standard

3. [R3]
   Are AI-assisted Development Tools Immune to Prompt Injection?

   preprint · original-paper

Record

Publication Record

Recommended citation

Ma, Chris. “When Tool Metadata Becomes an Instruction.” ACE Research Note ACE-RN-2026-008, 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.