high-quality-content-writerlisted

# High Quality Content Writer Use this skill as a closed-loop content system: draft or clean text, score it with deterministic gates, evaluate it in isolation, then retry with concrete fix instructions until quality passes or the retry budget is exhausted. ## Operating Model Default to this loop: 1. Prepare an isolated workdir under `.tmp/high-quality-content-writer/<timestamp>/`. 2. For websites, always separate scoring from extraction: - For trope scoring, run `slop_score.py --url <url>` first so the result is calibrated against the public AI Vetter action when available. - For token-efficient content review/rewrite, extract compact markdown through Jina Reader and save it as `source.md`: - `curl -L -s 'https://r.jina.ai/http://https://example.com/path' -o source.md` - If Jina fails or returns an error page, fall back to direct fetch/HTML extraction. 3. Save pasted text or extracted markdown as `source.md` and the current candidate as `candidate.md`. 4. Run deterministic scoring: - `scripts/slop_score.py --file candidate.md --pretty` - `scripts/quality_gate.py --file candidate.md --pretty` 5. If available, run an isolated evaluator subagent using `references/evaluator-flow.md`. Give it only the candidate text, scoring JSON, user brief, and the skill path. 6. If any gate fails, run an isolated writer subagent using `references/writer-flow.md`. Give it `source.md`, `candidate.md`, previous findings, and fix instructions. Ask it to write the next `candida