> For the complete documentation index, see [llms.txt](https://docs.lpp-minduniverse.org/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.lpp-minduniverse.org/lingua-pactum-protocol-lpp-documentation/6.-intent-and-trajectory-governance/6.9-trajectory-composition-failure.md).

# 6.9 Trajectory Composition Failure

One of the strongest trajectory-level principles is:

> **LocalPlausible ⇏ TrajectoryAdmissible**

This means:

> The fact that each individual step appears locally plausible does not imply that the composed trajectory remains constitutionally admissible.

#### Local Plausibility

A local action may appear reasonable because:

* it follows logically from the prior step,
* it uses an available tool,
* it remains inside a narrow technical permission,
* or it appears useful for task completion.

Local plausibility is therefore a weak property.

#### Trajectory Admissibility

Trajectory admissibility asks whether the composed path remains compatible with:

* legitimate authority,
* Human Intent H,
* scope,
* delegation constraints,
* material-mutation rules,
* cumulative externality,
* and other Layer 0 conditions.

#### Composition Failure Example

Suppose:

```
A₁ is locally plausible
A₂ is locally plausible
A₃ is locally plausible
A₄ is locally plausible
```

It does not follow that:

```
A₁ ∘ A₂ ∘ A₃ ∘ A₄
```

is admissible.

The sequence may create:

* scope drift,
* semantic drift,
* privilege accumulation,
* principal substitution,
* target expansion,
* or unintended external consequence.

Thus:

```
∀k LocalPlausible(A_k)
```

does not establish:

```
TrajectoryAdmissible(T)
```

#### Why Per-Step Guardrails Are Insufficient

A system that evaluates only individual tool calls can miss the governance meaning created through composition.

This is especially important in:

* multi-agent systems,
* cyber automation,
* financial workflows,
* infrastructure management,
* long-horizon planning,
* and physical AI.

Trajectory-level governance adds memory of legitimate purpose to execution control.

***

###