# Evidence Pack This folder is for dissertation evidence that can be regenerated on demand. ## Quick Start Run from repo root: ```bash bash scripts/collect_evidence.sh ``` Optional custom run id: ```bash bash scripts/collect_evidence.sh 2026-04-16-baseline ``` Outputs are written to `evidence/runs//`. To autofill templates for an existing run: ```bash bash scripts/autofill_evidence_templates.sh ``` To capture UI screenshots for an existing run: ```bash bash scripts/capture_ui_screenshots.sh ``` To include screenshot capture during evidence collection: ```bash KEYWARDEN_EVIDENCE_CAPTURE_SCREENSHOTS=1 bash scripts/collect_evidence.sh ``` ## What Is Collected - Git and environment snapshot (`meta/`) - Docker Compose state and container runtime info (`runtime/`) - Django test execution output (`tests/`) - Coverage summary + XML when available (`tests/coverage-report.txt`, `tests/coverage-xml.xml`) - Log extracts and endpoint/status summaries (`logs/`) - Derived operational metrics and summary (`evaluation/`) - RBAC/role-claim grep evidence (`docs/`) - Copy of diagram sources and report templates (`diagrams/`, `templates/`) - Autofilled template versions in each run (`templates/*.md`) - Optional automated UI screenshots (`screenshots/`) ## Operational Metrics The collector now derives evaluation-friendly metrics from `agent-access` logs: - heartbeat intervals per server - endpoint/status distributions per server - heartbeat error-to-recovery timings - markdown summary for direct report inclusion Artifacts are written to `evidence/runs//evaluation/`. ## Diagram Sources Mermaid sources live in `evidence/diagrams/`: - `architecture.mmd` - `deployment.mmd` - `erd.mmd` - `access-sync-flow.mmd` These are plain text and version-controlled, so they are easy to revise and re-render. ## Manual Evidence Templates Templates for remaining manual work are in `evidence/templates/`: - browser test matrix - security test report - standards checklist (OWASP/NIST-oriented)