Chapter 48 — Evidence dossier
Lesson map
Lesson 48 published. Map, example, EX142.
flowchart TD
N0["Learning outcomes"]
N1["48.1 Purpose"]
N2["48.2 Structure"]
N3["48.3 Claim–evidence links"]
N4["48.4 Sensitive content"]
N5["48.5 Integrity"]
N6["48.6 Completeness check"]
N7["48.7 Format"]
N0 --> N1 --> N2 --> N3 --> N4 --> N5 --> N6 --> N7
N7 --> EX["Worked example"]
EX --> Q["Exercise EX142"]
Logical drawing
flowchart LR
N0["Learning outcomes"]
N1["48.1 Purpose"]
N2["48.2 Structure"]
N3["48.3 Claim–evidence links"]
N4["48.4 Sensitive content"]
N5["48.5 Integrity"]
N6["48.6 Completeness check"]
N7["48.7 Format"]
N0 --> N1 --> N2 --> N3 --> N4 --> N5 --> N6 --> N7 --> Q["EX142"]
[Learning outcomes]
|
v
[48.1 Purpose]
|
v
[48.2 Structure]
|
v
[48.3 Claim–evidence links]
|
v
[48.4 Sensitive content]
|
v
[48.5 Integrity]
|
v
[48.6 Completeness check]
|
v
[48.7 Format]
|
v
[Worked example] --> [Exercise EX142]
Learning outcomes
You will (1) assemble an evidence dossier proving the capstone meets its requirements; (2) organise evidence for assessors and auditors; (3) protect sensitive content; (4) link every claim to its evidence.
48.1 Purpose
The dossier lets someone who was not there check every claim. Assessors use it to grade; Lindqvist keeps it for auditors and future engineers.
Claims in a report are only as strong as the evidence behind them. The dossier gathers that evidence in one place so an assessor, auditor or future engineer can check any statement without asking Amara. It also protects Lindqvist: when someone later asks how the depot detection was proved, the answer is on file. Amara builds it as the project runs.
48.2 Structure
Requirements table with status; design documents and review log; implementation records (commits, change tickets); verification results (unit, pipeline, timed tests); validation results (purple-team re-test); reliability results (completeness, failure trial); operational data (false alerts per week); remaining gaps.
The order follows the project from need to operation, so a reader can trace how each stage built on the last. Including remaining gaps alongside successes shows a complete picture. Operational data, such as false alerts per week, show how the system behaves in normal use, not just in tests.
48.3 Claim–evidence links
Each claim in the report points to a dossier item by ID: "Latency met (E-14: timed test results, 5 runs)". Unsupported claims are removed.
An ID lets a reader jump straight to the supporting item. Stating what the item contains, such as five runs, tells them how strong the evidence is. Removing unsupported claims keeps the report honest: if Amara cannot point to evidence, the claim does not belong in the report. Amara builds the link table before writing the conclusions, so the conclusions cannot claim more than the table supports.
48.4 Sensitive content
Host names, depot addresses and rule details that would help evasion are replaced with placeholders in the assessor copy; the full version stays inside Lindqvist with access control.
Details that would help someone evade detection, such as exact rule conditions, should not circulate beyond those who need them. Placeholders keep the assessor copy readable while removing risk. The full version stays under access control inside Lindqvist, where engineers and auditors with a need can consult it. The privacy officer checks the assessor copy.
48.5 Integrity
Exported files are listed with hashes so later readers can confirm they are unchanged.
A hash list lets anyone recompute each file's hash and compare it with the recorded value. A match shows the file has not changed since export, whether by accident or deliberately. The hash list itself is signed off by Paolo, tying it to a known point in time. Assessors receive the hash list separately from the files. Recomputing a few hashes takes minutes and gives them confidence that what they are grading is exactly what was submitted, with no later edits.
48.6 Completeness check
Paolo checks that each requirement has at least one verification and, where relevant, one validation item.
A requirement with no verification item has not been shown to work; one with no validation item, where validation applies, has not been shown to matter in practice. Paolo's check catches gaps before submission. Any missing items are either added or the requirement's status is marked as unproven. He also samples five claims at random and follows each to its evidence. Two links point to outdated file names, which Amara corrects before the dossier is sealed.
48.7 Format
Plain formats — PDF for documents, CSV for data, text for scripts — keep the dossier readable for years.
Proprietary formats may need software that is unavailable years later. PDF, CSV and plain text open in many tools on any system. Choosing them means the dossier can still be read when Lindqvist's current tools have been replaced, which matters for audits that may come much later. Amara converts a spreadsheet with formulas into CSV plus a short note explaining how each derived column was calculated, so nothing depends on the spreadsheet program.
Reading order
The dossier opens with a one-page index mapping each requirement to its evidence items, so an assessor can start from any requirement and reach the proof in two steps. A short glossary explains internal terms such as depot zone and possible-takeover alert. Amara asks a fellow student unfamiliar with Lindqvist to find three pieces of evidence using only the index; both difficulties she reports lead to clearer item titles.
Worked example
Paolo's check finds requirement "completeness ≥ 99.9%" supported only by one week of reconciliation. Amara adds the full four-week reconciliation export as E-22 and updates the requirements table.
Synthetic case. The dossier is invented.
Errors and limits
Dossiers record what was tested; untested behaviour remains uncertain.
Exercises
- EX142 — Dossier sections. List five dossier sections. Rubric: 3 = five of requirements, design/review, implementation, verification, validation, reliability, operations, gaps; 2 = three; 1 = one; 0 = none.
- EX143 — Claim links. How are claims linked to evidence? Rubric: 3 = each claim cites dossier item ID; unsupported removed; 2 = partial; 1 = vague; 0 = none.
- EX144 — Completeness gap. What gap did Paolo find, and how was it fixed? Rubric: 3 = completeness backed by one week only; full four-week export added; 2 = partial; 1 = vague; 0 = none.
Synthesis
A dossier links every claim to evidence, protects sensitive detail and lets others check the work.
Continue studying — chapter 48 of 50
← Chapter 47 (draft) · Course chapter index · Chapter 49 (draft) →