Fresh sustainability audit reproduction

Use the web-sustainability-agents repository and its locked uv dependencies.
This archive preserves original collector inputs, skills, three specialist
results, acceptance receipts, coordinator resolutions, final judgments and
editorial inputs. Original task paths record the execution environment.
Native model tool transcripts are not included; actual model/task settings
and output digests are recorded in the receipts.

Extract into a new directory. Recalculate without network or model calls:

uv run --locked --script /path/to/repo/audit/assess.py --reviews session/reviews.json --rubric session/rubric.json --weights session/impact-weights.json --assets session --output recalculated.json

Render with the repository's renderer (use MkDocs for themed presentation):

uv run --locked --script /path/to/repo/audit/render_review.py --session session --artifact-prefix session --editorial presentation/editorial.json --screenshots presentation/screenshots/manifest.json --image-prefix presentation/screenshots --output report.md

For a new review, create new task files pointing at the extracted frozen session
with audit/cli.py tool orchestrate tasks. Do not edit existing accepted results.
For new measurements, copy/refine the plan and create a new collection directory.
Adjust source roots to the matching supplied repository revisions.

Comparability: scoped judgments are not whole-guideline conformance. Live pages
and model judgments may change. No annual scenario or hardware-energy result
was produced in this run. Later illustrations do not prove earlier measurements.
