algal-writing

Use writing-audit before spending model context rereading a whole draft.

bunx algal-skills run writing-audit --args '{"src":{"path":"README.md","cwd":"."}}'
bunx algal-skills run writing-evaluate --args @writing-args.json

writing-audit is deterministic and makes zero model calls. writing-evaluate adds one typed decide effect for dominant issue, publish-readiness probability, and a 1–5 quality score. It evaluates but never rewrites prose. When Jev is unavailable, use the deterministic audit and edit with the active agent.