PAT-0140
Boundary Leakage
A structural condition where effects, permissions, state, data, or authority cross a declared boundary without an authorized exception.
Primary Lenses
Isolation Boundary Lens
Evaluates whether structural constraints, effects, and regions remain contained within declared boundaries.
Lens Application
The Isolation Boundary Lens is a strong inspection mechanism for Boundary Leakage because it focuses attention on whether declared isolation regions actually constrain movement of effects, permissions, state, data, and authority. It helps inspect whether any cross-boundary behavior is explicitly authorized.
Inspect For
- Effects escaping declared regions
- Shared state or data crossing boundaries
- Permission or authority bleed-through
- Missing or implicit boundary exceptions
Avoid
Treat cross-boundary behavior as an inspection signal, not proof that Boundary Leakage is present.
Propagation Lens
Traces how structural declarations, effects, or state changes propagate across boundaries or stages.
Lens Application
The Propagation Lens is a strong inspection mechanism for Boundary Leakage because it follows effects, permissions, state, data, and authority as they move across boundaries, stages, or scopes. It helps surface where propagation exceeds the intended scope, skips an exception path, or carries authority into places not authorized by the boundary.
Inspect For
- Effects crossing boundary edges without an explicit exception
- Permissions or authority persisting beyond intended scope
- State or data copied into downstream stages unexpectedly
- Boundary checks that occur after propagation has already happened
Avoid
Treat propagation evidence as a signal for closer inspection, not confirmation that Boundary Leakage has occurred.
Secondary Lenses
Boundary Compliance Lens
Evaluates observed structure against declared boundary posture, including allow, block, and exception rules.
Lens Application
The Boundary Compliance Lens applies indirectly to Boundary Leakage by checking whether effects, permissions, state, data, or authority cross declared boundaries in ways that match the stated boundary posture. It helps surface crossings that lack allow rules, appear to bypass block rules, or require exception review.
Inspect For
- Cross-boundary flows without matching allow rules
- Permissions or authority extending beyond declared scope
- State or data movement relying on unclear exceptions
- Block rules bypassed by informal paths
Avoid
Treat boundary posture mismatches as review signals, not proof that Boundary Leakage has occurred.
Primary Issue Matches
Agent Modified Unrelated State
An agent changes state outside the requested task, target, file, record, workflow, or authorized scope.
AI Uses External Information When Forbidden
The AI uses outside knowledge, sources, tools, memory, or assumptions after the task forbids external information or limits the allowed source set.
Context Leaks Between Tasks
Context, assumptions, constraints, examples, files, or decisions from one task affect another task where they should not apply.
File-Bounded Task Uses Outside Content
The AI is asked to work only from a specific file or document but uses content, assumptions, or sources outside that file.
Retrieval Exceeds Evidence Limit
The AI retrieves, uses, cites, or considers more evidence than the task permits or more than the review surface can support.
Review Outcome Changes Unrelated Environment
A review result, approval, rejection, or classification changes state outside the environment, case, file, or workflow it was meant to govern.
Supporting Issue Matches
Action Changed Something Else Too
An AI or agent action makes the requested change but also changes another object, field, file, state, rule, or workflow element that was not supposed to change.
AI Memory Has No Governance
Saved or persistent AI memory affects output without clear rules for ownership, scope, review, update, expiry, or removal.
AI Memory Updated Without Asking
AI memory, saved context, preference, or durable state is updated without the user clearly asking for or approving that update.
AI Touches Unrelated Scope
The AI affects, edits, analyzes, changes, or reasons over material outside the scope of the requested task.
Hidden Rule Overrides Visible Instruction
A hidden, upstream, system, policy, tool, or product rule changes or overrides the visible instruction the user expects the AI to follow.
Local Exception Grows Into Policy
A local exception, special case, or one-off allowance begins to function like a general policy.
Local Rule Spreads to Broader Cases
A rule intended for one local case, file, context, user, workflow, or exception begins affecting broader cases.
Policy Exception Spreads Too Far
A narrow policy exception, allowance, or special case spreads beyond its intended scope and begins governing broader cases.
Prompt Does Not Say What to Exclude
The prompt declares what to include but does not declare what should be excluded, allowing unwanted scope, sources, content, or actions into the result.
Routing Overrides Task Intent
Routing, mode selection, agent behavior, or workflow classification sends the task down a path that overrides what the user was trying to accomplish.