# Completed demonstration example — embargoed observatory permission audit

**Authored alternate example.** Read this local research embargo before deciding any permission.

## Published authority
Reader: read research only. Maintainer: service equipment only; embargoed research is not readable. Custodian: read research and approve release. Every other combination is denied.

## Preserved faulty results
| Request | Expected | Observed before repair |
| --- | --- | --- |
| Maintainer / read | Deny | Allow |
| Reader / read | Allow | Deny |

The other seven decisions already matched. The first failure exposed an embargoed resource; the second blocked an authorised reader.

## Version 1 — repaired charter
| Role | Read | Service | Approve |
| --- | --- | --- | --- |
| Reader | Allow | Deny | Deny |
| Maintainer | Deny | Allow | Deny |
| Custodian | Allow | Deny | Allow |

All nine cells were compared with the charter. Positive checks preserved reader reading, maintainer service and custodian reading/approval. Negative checks retained the embargo and remaining limits. The repair changed only maintainer/read and reader/read.

## Version 2 — new release hold
A later signed notice suspended custodian approval. The other eight decisions stayed as in Version 1. Reader and custodian could still read; maintainer could still service but could not read research. The previous correct approval is preserved under its earlier authority.

## Reflection
I initially expected equipment responsibility to include research visibility. The embargo contradicted that shortcut. Checking forbidden operations alone would also have missed the blocked reader, so I recorded both kinds of failure. The release hold was a new rule, not proof that the original approval test was dishonest. In the assigned lab I must derive a fresh table from its own charter and identify exactly which right its later notice changes. I can reuse the testing method without reusing these permission answers.

Named output: permission-audit. These are fictional local decisions, not an external security test.

## Method to carry forward

Derive expected permissions from a charter before testing observed behaviour.

The assigned role policy grants different read rights. Copying this completed matrix would apply the wrong authority.

The complete example explains one defensible approach. In your own challenge, select observations and tests from the supplied rules, keep your prediction separate from the result, and justify both the checks you perform and those you omit.

The chapter investigation and its explained record complete the core method. The scene delivery is an optional extension; it is not evidence of deeper understanding on its own.