feat(fetch_job_links): make big-tech/non-tech filters fully optional

fix(fetch_job_links): add 1-3s random delay between searches to avoid rate limiting

feat: public release — configurable filters, single-subagent summarize, README, INSTALL_AGENT

refactor: merge jd-rater into internship-scout, remove lobster pipeline and mcp_call, add summarize_jds.py to skill

refactor: replace chrome-mcp JD fetch with osascript DOM extractor in internship-scout

feat: internship-scout 重建 — agentTurn 隔离提取 + fetch_jd_dom 硬阻断 + 幻觉检测

revert: internship-scout 回退到 e93e49b 状态（删除 lobster pipeline 及后续脚本）

feat: add lobster pipeline and deterministic scout scripts (no llm-task plugin)

refactor: switch scout workflow to llm-task orchestration (JD fetch/summary/notion sync) with batch<=5

docs: add internship-scout initialization checklist before batch run

feat: dual-write jd_summary+jd_full to notion and switch summary spec to 30-50 chars

refactor: scout JD flow via live DOM and keep jd_summary(20 chars)+jd_full

refactor: store full jd in yaml (no summary) and sync jd_full to notion

feat: add jd-rater skill with multi-dimensional scoring; fix jd_quality for 7 entries; decouple rating from collection

fix(internship-scout): use PyYAML to parse internships.yaml instead of broken regex