[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M77N63WBJ9SCPHHFCRWZ681R":3,"skill-same:01M77N63WBJ9SCPHHFCRWZ681R":81},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":30,"install":31,"resources":44},"01M77N63WBJ9SCPHHFCRWZ681R","news-rss-watch",null,"Polls a seeded, deterministic RSS\u002FAtom feed registry covering US news, China news, Singapore news, global market news, important\u002Fworld news, and arXiv research feeds — with per-feed watermark dedup so each run surfaces ONLY new items. Use whenever the user asks to \"check the news\", \"poll feeds\", \"what's new since yesterday\", \"morning news digest\", \"watch this feed\", \"seed\u002Ftrack RSS feeds\", wants a comprehensive US+China+Singapore news outlook, or wants new arXiv papers on a finance\u002FML topic watched. Triggers on phrases like \"news RSS\", \"add this feed\", \"what's new in the feeds\", \"daily news brief\", \"track arxiv papers on X\". This is the deterministic data layer — pair with `daily-market-watch` when the user wants a synthesized, source-cited market report rather than raw new items.","kangcodex\u002Ffinance-skills","kangcodex","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F97664140?v=4","01M77N5RD8RGSEX3HEPW28G3TT","Kang","01M77N5RDVSD75TRTA7DFSJMHN",0,"MIT","https:\u002F\u002Fgithub.com\u002Fkangcodex\u002Ffinance-skills\u002Fblob\u002Fmain\u002Fskills\u002Fnews-rss-watch\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: news-rss-watch\ndescription: >-\n  Polls a seeded, deterministic RSS\u002FAtom feed registry covering US news,\n  China news, Singapore news, global market news, important\u002Fworld news, and\n  arXiv research feeds — with per-feed watermark dedup so each run surfaces\n  ONLY new items. Use whenever the user asks to \"check the news\", \"poll\n  feeds\", \"what's new since yesterday\", \"morning news digest\", \"watch this\n  feed\", \"seed\u002Ftrack RSS feeds\", wants a comprehensive US+China+Singapore\n  news outlook, or wants new arXiv papers on a finance\u002FML topic watched.\n  Triggers on phrases like \"news RSS\", \"add this feed\", \"what's new in the\n  feeds\", \"daily news brief\", \"track arxiv papers on X\". This is the\n  deterministic data layer — pair with `daily-market-watch` when the user\n  wants a synthesized, source-cited market report rather than raw new items.\n---\n\n# news-rss-watch\n\nDeterministic, seedable news feed watcher. Polls a config-driven feed\nregistry (`feeds.json`) with watermark dedup — each run emits only items\npublished since the last run, as strict JSON. No LLM interpretation in\nthe fetch loop: the agent runs the script and renders the output.\n\n## Works with any agent\n\nThis skill is agent-agnostic by construction:\n\n- **Pure CLI + JSON contract.** The scripts take flags, print strict\n  JSON\u002Fmarkdown to stdout, and exit with a status code. No agent\n  framework APIs, no tool hooks, no environment variables beyond\n  `NEWS_WATCHER_STATE_DIR` \u002F `NEWS_WATCHER_DIGEST_DIR` (both optional).\n- **Only needs a shell.** Any agent with a terminal tool (Claude Code,\n  Codex, Cursor, OpenCode, Gemini CLI, plain cron, a human) can drive\n  it — the SKILL.md is instructions for whatever agent is running it.\n- **Stdlib-only Python.** No pip installs, no virtualenv, no network\n  calls beyond fetching the feeds themselves.\n- **Deterministic output.** Identical inputs → identical JSON, so any\n  agent parses the same contract. The skill never depends on the\n  agent's memory or prior turns.\n\n## When to use\n\n- \"Check the news\" \u002F \"what's new since yesterday\" \u002F \"poll the feeds\"\n- \"Daily news digest\" for US + China + Singapore + markets + world\n- \"Add this feed \u002F track this website\" — deterministic `--action add`\n- \"What feeds are you tracking?\" — `--action list`\n- \"Watch arXiv for new papers on \u003Ctopic>\" — `scripts\u002Fwatch_arxiv.py`\n- Any cron \u002F scheduled briefing job: run the script, parse JSON, stay\n  silent when the item count is 0.\n\n## Mental model\n\n1. `feeds.json` — the seed registry. Pre-seeded with **live-verified**\n   feeds (checked 2026-08-08). Agent never edits feeds by hand; it uses\n   `--action add` \u002F `--action list` so the registry stays deterministic.\n2. `scripts\u002Fwatch_feeds.py` — batch poller. One command, all feeds,\n   one JSON result. Per-feed isolation: a broken feed is reported in\n   `errors`, never crashes the batch.\n3. Watermark state — `~\u002F.finance-skills\u002Fwatcher-state\u002Fnewsrss_\u003Cid>.json`\n   (override with `NEWS_WATCHER_STATE_DIR`). First run = baseline:\n   records IDs, emits nothing. This prevents flooding the agent's\n   context with old articles on the first poll.\n4. `scripts\u002Fwatch_arxiv.py` — same pattern against the arXiv Atom API\n   (one-shot arXiv search turned into a watcher).\n\n## Feed registry (seeded categories)\n\n| Category | Focus | Seed feeds |\n| --- | --- | --- |\n| `us` | US news + policy + regulators | NPR, CNBC, WSJ Markets, SEC press, Fed press, Google News US |\n| `china` | China news (first-party + HK) | SCMP news\u002Fbusiness, CGTN, Xinhua English, Google News CN (zh) |\n| `singapore` | SG local news + markets | CNA, Straits Times, Business Times SG, Google News SG, SG business topic |\n| `markets` | Global market news | MarketWatch, Yahoo Finance, FT, BBC Business, Investing.com, CoinDesk, Cointelegraph, Google Business |\n| `important` | World \u002F headline news | BBC World, Guardian World\u002FBusiness, Google News World |\n| `research` | arXiv finance + AI | q-fin.ST\u002FEC\u002FTR\u002FPM\u002FMF\u002FRM\u002FGN, econ.GN, cs.AI (weekday-only updates) |\n\nRegion-source rule (mirrors ADR-002): China category uses Chinese\nfirst-party \u002F HK sources (Xinhua, CGTN, SCMP) — not US outlets covering\nChina. Singapore uses local outlets only. See\n`references\u002Ffeed-registry.md` for verification notes and feeds that\nwere tested and rejected (dead, paywalled, rate-limited).\n\n## Workflow\n\n```\nNews poll run:\n- [ ] 1) Pick scope: all categories, or --categories us,china,...\n- [ ] 2) Run: python scripts\u002Fwatch_feeds.py [--categories ...] [--max N] [--budget N]\n- [ ] 3) Parse the JSON. count == 0 → reply \"No new updates.\" Do NOT fabricate news.\n- [ ] 4) Render stories grouped by category, with title + URL + score.\n- [ ] 5) On user request, synthesize: use daily-market-watch for the full report.\n```\n\n### Polling (the default action)\n\n```bash\npython scripts\u002Fwatch_feeds.py\n# → JSON: {\"count\": N, \"first_run\": bool, \"results\": [{category, clusters: [...]}], \"errors\": [...]}\n```\n\nClustering is ON by default (anti-noise): the same story picked up by N\nfeeds becomes ONE story. Each cluster:\n\n```json\n{\n  \"cluster_id\": 0, \"score\": 3.5, \"matched_keywords\": [\"fed\", \"cpi\"],\n  \"category\": \"markets\", \"feed_id\": \"marketwatch_top\",\n  \"feed_name\": \"MarketWatch Top Stories\",\n  \"cluster_size\": 2, \"other_sources\": [\"Google News Business\"],\n  \"items\": [{\"id\", \"title\", \"url\", \"summary\", \"published\"}]\n}\n```\n\n- **score** = source weight (3 = regulator\u002Ffirst-party, 2 = major outlet,\n  1 = aggregator) + 0.5 per importance keyword hit (fed, cpi, tariff,\n  mas, pboc, war, earnings, …). Sort by score when rendering.\n- **Attribution**: the highest-weight source wins the category + link;\n  `other_sources` lists the rest. A China story covered by Xinhua + BBC\n  stays in `china`, not `important`.\n- **Budget**: `--budget N` caps stories per category (default 8,\n  `0` = unlimited).\n\nBehavior flags:\n\n- First ever run records a baseline and returns `count: 0` with\n  `first_run: true` — tell the user the watcher is primed, then re-run\n  (or use `--fresh` for a full digest).\n- `--categories us,markets` — scope to one or more categories.\n- `--fresh` — emit everything, ignoring watermarks (still records state\n  so the next normal poll stays quiet).\n- `--max 15` — per-feed cap (Google News feeds default to 15).\n- `--with-summary` — include a ≤500-char summary snippet per item.\n- `--no-cluster` — flat per-feed view (debugging single feeds; no dedup).\n- `--format markdown` — machine-shaped markdown for cron delivery.\n\n### Daily digest (cron-friendly)\n\n```bash\npython scripts\u002Fwatch_feeds.py --action digest --max 5 --budget 5\n# → writes ~\u002F.finance-skills\u002Fdigests\u002Fnews-digest-YYYY-MM-DD.md (override\n#   dir with NEWS_WATCHER_DIGEST_DIR), prints {\"digest\": path, \"stories\": N}\n```\n\nFresh full poll → cluster → per-category markdown file. Zero LLM needed:\npoint a nightly cron at it and ship the file to your inbox\u002Fchannel.\n\n### Adding a feed (deterministic, no file editing)\n\n```bash\npython scripts\u002Fwatch_feeds.py --action add --name \"my_feed\" --url \"https:\u002F\u002Fexample.com\u002Frss\" --category us\n```\n\nValidation: `--category` must be one of\n`us, china, singapore, markets, important, research`. Duplicate\nurl\u002Fid → hard error, no mutation.\n\n### Search-as-feed (gap filler)\n\n```bash\n# Any topic\u002Foutlet Google News can cover — e.g. MAS has no RSS:\npython scripts\u002Fwatch_feeds.py --action add-search \\\n  --query \"MAS Singapore monetary policy\" --category singapore \\\n  --hl en-SG --gl SG --ceid SG:en\n```\n\nBuilds a Google News search RSS feed and adds it like any other feed.\nRegion-scoped via `--hl\u002F--gl\u002F--ceid` (defaults en-US\u002FUS\u002FUS:en).\n\n### Checking \u002F resetting\n\n```bash\npython scripts\u002Fwatch_feeds.py --action list                 # show registry\npython scripts\u002Fwatch_feeds.py --action check                # verify every feed URL + XML\npython scripts\u002Fwatch_feeds.py --action reset --feed-id cna_main   # clear one watermark\npython scripts\u002Fwatch_feeds.py --action reset                # clear all watermarks\n```\n\n### Watching arXiv topics (research category)\n\n```bash\npython scripts\u002Fwatch_arxiv.py --query \"cat:q-fin.ST AND agent\" --name qfin_agents\npython scripts\u002Fwatch_arxiv.py --query \"ti:LLM trading\" --max 5\npython scripts\u002Fwatch_arxiv.py --query \"all:deep learning volatility\" --fresh\n```\n\nEach run emits only papers new since the last run (watermark keyed on\narXiv id). Exactly 1 API request per run — safe for cron. Supports the\nfull arXiv query syntax (`all:`, `ti:`, `abs:`, `au:`, `cat:`, AND\u002FOR).\n\n## Guardrails\n\n- **Never fabricate news.** If `count` is 0, respond with no-new-updates.\n  If a feed errored, report it from `errors` — never substitute another\n  feed's items.\n- **Never edit `feeds.json` by hand.** Use `--action add` \u002F `add-search`.\n  This keeps state deterministic and auditable.\n- **Never rename a feed id mid-stream.** The watermark is keyed on the\n  id; renaming resets dedup and floods context with old items.\n- **First-run semantics are a feature.** Baseline recording prevents\n  replay. Use `--fresh` only when the user explicitly wants a full\n  digest.\n- **Clustering is deterministic, not editorial.** Two stories are merged\n  only when title tokens overlap ≥60% after stopword\u002Foutlet stripping.\n  Do not hand-merge or hand-split clusters when rendering; show what\n  the script emits. If clustering looks wrong (over-merge of distinct\n  stories), report it — the threshold lives in the script.\n- **Score is a sort key, not a verdict.** Higher score = more keywords +\n  more authoritative source. Render in score order but never drop a\n  story the user asked about just because it scored low.\n- **Per-feed failure ≠ batch failure.** One dead feed goes to `errors`;\n  the rest still poll. Report errors to the user, don't retry endlessly\n  (respect HTTP 429 by backing off — the next scheduled poll retries).\n- **High-churn feeds are capped.** Google News feeds emit dozens of\n  items; `max_per_poll` in feeds.json caps them at 15, `--budget` caps\n  stories per category.\n- **Don't re-order or re-parse feeds yourself.** The script is the\n  single source of truth; if a feed breaks, run `--action check`.\n\n## Tests\n\n```bash\npython3 -m unittest discover -s skills\u002Fnews-rss-watch\u002Ftests\n```\n\n27 offline tests: RSS\u002FAtom parsing (incl. malformed-CDATA repair), title\nnormalization, clustering, importance scoring, budget caps, watermark\nbaseline\u002Fdedup\u002Fcorruption, feed-add validation, search-URL building,\ndigest writing, arXiv Atom parsing. No network needed.\n\n## Companion skills\n\n- `daily-market-watch` — when the user wants a synthesized, source-cited\n  market report (6 sections, region zoom) instead of raw new items: poll\n  first with this skill, then hand the new items to daily-market-watch\n  as the news layer.\n- `smart-money-tracker` — when a news item mentions a 13F\u002FSTOCK Act\n  filing, cross-check with the tracker.\n\n## References\n\n- `references\u002Ffeed-registry.md` — how feeds were verified, what was\n  rejected and why, notes on flaky feeds (arXiv weekend silence, SEC\n  User-Agent requirement, Reddit rate limits, MAS has no working RSS).\n- `feeds.json` — the live seed registry (edit via `--action add` only).\n\n## Example\n\n- `examples\u002Fsample-report.md` — a real clustered poll run (all 5 news\n  categories, scores, dedup evidence). Golden output for the eval set;\n  use it as the shape reference when rendering.\n- `examples\u002Fsample-digest.md` — a real nightly digest file\n  (`--action digest`). Use as the shape reference for cron delivery.\n","# news-rss-watch\n\nDeterministic, seedable news feed watcher. Polls a config-driven feed\nregistry (`feeds.json`) with watermark dedup — each run emits only items\npublished since the last run, as strict JSON. No LLM interpretation in\nthe fetch loop: the agent runs the script and renders the output.\n\n## Works with any agent\n\nThis skill is agent-agnostic by construction:\n\n- **Pure CLI + JSON contract.** The scripts take flags, print strict\n  JSON\u002Fmarkdown to stdout, and exit with a status code. No agent\n  framework APIs, no tool hooks, no environment variables beyond\n  `NEWS_WATCHER_STATE_DIR` \u002F `NEWS_WATCHER_DIGEST_DIR` (both optional).\n- **Only needs a shell.** Any agent with a terminal tool (Claude Code,\n  Codex, Cursor, OpenCode, Gemini CLI, plain cron, a human) can drive\n  it — the SKILL.md is instructions for whatever agent is running it.\n- **Stdlib-only Python.** No pip installs, no virtualenv, no network\n  calls beyond fetching the feeds themselves.\n- **Deterministic output.** Identical inputs → identical JSON, so any\n  agent parses the same contract. The skill never depends on the\n  agent's memory or prior turns.\n\n## When to use\n\n- \"Check the news\" \u002F \"what's new since yesterday\" \u002F \"poll the feeds\"\n- \"Daily news digest\" for US + China + Singapore + markets + world\n- \"Add this feed \u002F track this website\" — deterministic `--action add`\n- \"What feeds are you tracking?\" — `--action list`\n- \"Watch arXiv for new papers on \u003Ctopic>\" — `scripts\u002Fwatch_arxiv.py`\n- Any cron \u002F scheduled briefing job: run the script, parse JSON, stay\n  silent when the item count is 0.\n\n## Mental model\n\n1. `feeds.json` — the seed registry. Pre-seeded with **live-verified**\n   feeds (checked 2026-08-08). Agent never edits feeds by hand; it uses\n   `--action add` \u002F `--action list` so the registry stays deterministic.\n2. `scripts\u002Fwatch_feeds.py` — batch poller. One command, all feeds,\n   one JSON result. Per-feed isolation: a broken feed is reported in\n   `errors`, never crashes the batch.\n3. Watermark state — `~\u002F.finance-skills\u002Fwatcher-state\u002Fnewsrss_\u003Cid>.json`\n   (override with `NEWS_WATCHER_STATE_DIR`). First run = baseline:\n   records IDs, emits nothing. This prevents flooding the agent's\n   context with old articles on the first poll.\n4. `scripts\u002Fwatch_arxiv.py` — same pattern against the arXiv Atom API\n   (one-shot arXiv search turned into a watcher).\n\n## Feed registry (seeded categories)\n\n| Category | Focus | Seed feeds |\n| --- | --- | --- |\n| `us` | US news + policy + regulators | NPR, CNBC, WSJ Markets, SEC press, Fed press, Google News US |\n| `china` | China news (first-party + HK) | SCMP news\u002Fbusiness, CGTN, Xinhua English, Google News CN (zh) |\n| `singapore` | SG local news + markets | CNA, Straits Times, Business Times SG, Google News SG, SG business topic |\n| `markets` | Global market news | MarketWatch, Yahoo Finance, FT, BBC Business, Investing.com, CoinDesk, Cointelegraph, Google Business |\n| `important` | World \u002F headline news | BBC World, Guardian World\u002FBusiness, Google News World |\n| `research` | arXiv finance + AI | q-fin.ST\u002FEC\u002FTR\u002FPM\u002FMF\u002FRM\u002FGN, econ.GN, cs.AI (weekday-only updates) |\n\nRegion-source rule (mirrors ADR-002): China category uses Chinese\nfirst-party \u002F HK sources (Xinhua, CGTN, SCMP) — not US outlets covering\nChina. Singapore uses local outlets only. See\n`references\u002Ffeed-registry.md` for verification notes and feeds that\nwere tested and rejected (dead, paywalled, rate-limited).\n\n## Workflow\n\n```\nNews poll run:\n- [ ] 1) Pick scope: all categories, or --categories us,china,...\n- [ ] 2) Run: python scripts\u002Fwatch_feeds.py [--categories ...] [--max N] [--budget N]\n- [ ] 3) Parse the JSON. count == 0 → reply \"No new updates.\" Do NOT fabricate news.\n- [ ] 4) Render stories grouped by category, with title + URL + score.\n- [ ] 5) On user request, synthesize: use daily-market-watch for the full report.\n```\n\n### Polling (the default action)\n\n```bash\npython scripts\u002Fwatch_feeds.py\n# → JSON: {\"count\": N, \"first_run\": bool, \"results\": [{category, clusters: [...]}], \"errors\": [...]}\n```\n\nClustering is ON by default (anti-noise): the same story picked up by N\nfeeds becomes ONE story. Each cluster:\n\n```json\n{\n  \"cluster_id\": 0, \"score\": 3.5, \"matched_keywords\": [\"fed\", \"cpi\"],\n  \"category\": \"markets\", \"feed_id\": \"marketwatch_top\",\n  \"feed_name\": \"MarketWatch Top Stories\",\n  \"cluster_size\": 2, \"other_sources\": [\"Google News Business\"],\n  \"items\": [{\"id\", \"title\", \"url\", \"summary\", \"published\"}]\n}\n```\n\n- **score** = source weight (3 = regulator\u002Ffirst-party, 2 = major outlet,\n  1 = aggregator) + 0.5 per importance keyword hit (fed, cpi, tariff,\n  mas, pboc, war, earnings, …). Sort by score when rendering.\n- **Attribution**: the highest-weight source wins the category + link;\n  `other_sources` lists the rest. A China story covered by Xinhua + BBC\n  stays in `china`, not `important`.\n- **Budget**: `--budget N` caps stories per category (default 8,\n  `0` = unlimited).\n\nBehavior flags:\n\n- First ever run records a baseline and returns `count: 0` with\n  `first_run: true` — tell the user the watcher is primed, then re-run\n  (or use `--fresh` for a full digest).\n- `--categories us,markets` — scope to one or more categories.\n- `--fresh` — emit everything, ignoring watermarks (still records state\n  so the next normal poll stays quiet).\n- `--max 15` — per-feed cap (Google News feeds default to 15).\n- `--with-summary` — include a ≤500-char summary snippet per item.\n- `--no-cluster` — flat per-feed view (debugging single feeds; no dedup).\n- `--format markdown` — machine-shaped markdown for cron delivery.\n\n### Daily digest (cron-friendly)\n\n```bash\npython scripts\u002Fwatch_feeds.py --action digest --max 5 --budget 5\n# → writes ~\u002F.finance-skills\u002Fdigests\u002Fnews-digest-YYYY-MM-DD.md (override\n#   dir with NEWS_WATCHER_DIGEST_DIR), prints {\"digest\": path, \"stories\": N}\n```\n\nFresh full poll → cluster → per-category markdown file. Zero LLM needed:\npoint a nightly cron at it and ship the file to your inbox\u002Fchannel.\n\n### Adding a feed (deterministic, no file editing)\n\n```bash\npython scripts\u002Fwatch_feeds.py --action add --name \"my_feed\" --url \"https:\u002F\u002Fexample.com\u002Frss\" --category us\n```\n\nValidation: `--category` must be one of\n`us, china, singapore, markets, important, research`. Duplicate\nurl\u002Fid → hard error, no mutation.\n\n### Search-as-feed (gap filler)\n\n```bash\n# Any topic\u002Foutlet Google News can cover — e.g. MAS has no RSS:\npython scripts\u002Fwatch_feeds.py --action add-search \\\n  --query \"MAS Singapore monetary policy\" --category singapore \\\n  --hl en-SG --gl SG --ceid SG:en\n```\n\nBuilds a Google News search RSS feed and adds it like any other feed.\nRegion-scoped via `--hl\u002F--gl\u002F--ceid` (defaults en-US\u002FUS\u002FUS:en).\n\n### Checking \u002F resetting\n\n```bash\npython scripts\u002Fwatch_feeds.py --action list                 # show registry\npython scripts\u002Fwatch_feeds.py --action check                # verify every feed URL + XML\npython scripts\u002Fwatch_feeds.py --action reset --feed-id cna_main   # clear one watermark\npython scripts\u002Fwatch_feeds.py --action reset                # clear all watermarks\n```\n\n### Watching arXiv topics (research category)\n\n```bash\npython scripts\u002Fwatch_arxiv.py --query \"cat:q-fin.ST AND agent\" --name qfin_agents\npython scripts\u002Fwatch_arxiv.py --query \"ti:LLM trading\" --max 5\npython scripts\u002Fwatch_arxiv.py --query \"all:deep learning volatility\" --fresh\n```\n\nEach run emits only papers new since the last run (watermark keyed on\narXiv id). Exactly 1 API request per run — safe for cron. Supports the\nfull arXiv query syntax (`all:`, `ti:`, `abs:`, `au:`, `cat:`, AND\u002FOR).\n\n## Guardrails\n\n- **Never fabricate news.** If `count` is 0, respond with no-new-updates.\n  If a feed errored, report it from `errors` — never substitute another\n  feed's items.\n- **Never edit `feeds.json` by hand.** Use `--action add` \u002F `add-search`.\n  This keeps state deterministic and auditable.\n- **Never rename a feed id mid-stream.** The watermark is keyed on the\n  id; renaming resets dedup and floods context with old items.\n- **First-run semantics are a feature.** Baseline recording prevents\n  replay. Use `--fresh` only when the user explicitly wants a full\n  digest.\n- **Clustering is deterministic, not editorial.** Two stories are merged\n  only when title tokens overlap ≥60% after stopword\u002Foutlet stripping.\n  Do not hand-merge or hand-split clusters when rendering; show what\n  the script emits. If clustering looks wrong (over-merge of distinct\n  stories), report it — the threshold lives in the script.\n- **Score is a sort key, not a verdict.** Higher score = more keywords +\n  more authoritative source. Render in score order but never drop a\n  story the user asked about just because it scored low.\n- **Per-feed failure ≠ batch failure.** One dead feed goes to `errors`;\n  the rest still poll. Report errors to the user, don't retry endlessly\n  (respect HTTP 429 by backing off — the next scheduled poll retries).\n- **High-churn feeds are capped.** Google News feeds emit dozens of\n  items; `max_per_poll` in feeds.json caps them at 15, `--budget` caps\n  stories per category.\n- **Don't re-order or re-parse feeds yourself.** The script is the\n  single source of truth; if a feed breaks, run `--action check`.\n\n## Tests\n\n```bash\npython3 -m unittest discover -s skills\u002Fnews-rss-watch\u002Ftests\n```\n\n27 offline tests: RSS\u002FAtom parsing (incl. malformed-CDATA repair), title\nnormalization, clustering, importance scoring, budget caps, watermark\nbaseline\u002Fdedup\u002Fcorruption, feed-add validation, search-URL building,\ndigest writing, arXiv Atom parsing. No network needed.\n\n## Companion skills\n\n- `daily-market-watch` — when the user wants a synthesized, source-cited\n  market report (6 sections, region zoom) instead of raw new items: poll\n  first with this skill, then hand the new items to daily-market-watch\n  as the news layer.\n- `smart-money-tracker` — when a news item mentions a 13F\u002FSTOCK Act\n  filing, cross-check with the tracker.\n\n## References\n\n- `references\u002Ffeed-registry.md` — how feeds were verified, what was\n  rejected and why, notes on flaky feeds (arXiv weekend silence, SEC\n  User-Agent requirement, Reddit rate limits, MAS has no working RSS).\n- `feeds.json` — the live seed registry (edit via `--action add` only).\n\n## Example\n\n- `examples\u002Fsample-report.md` — a real clustered poll run (all 5 news\n  categories, scores, dedup evidence). Golden output for the eval set;\n  use it as the shape reference when rendering.\n- `examples\u002Fsample-digest.md` — a real nightly digest file\n  (`--action digest`). Use as the shape reference for cron delivery.\n",{"name":5,"description":7},"main","471b6a738d7a6da27eaa328e3fed4341337dd279","skills\u002Fnews-rss-watch",2,"https:\u002F\u002Fgithub.com\u002Fkangcodex\u002Ffinance-skills","https:\u002F\u002Fgithub.com\u002Fkangcodex","2026-09-13 17:24:23",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add kangcodex\u002Ffinance-skills --skill news-rss-watch",{"tool":37,"command":38},"bunx","bunx skills add kangcodex\u002Ffinance-skills --skill news-rss-watch",{"tool":40,"command":41},"pnpm","pnpm dlx skills add kangcodex\u002Ffinance-skills --skill news-rss-watch","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fkangcodex\u002Ffinance-skills\u002Fblob\u002Fmain\u002Fskills\u002Fnews-rss-watch\u002FSKILL.md",true,[45,51,55,59,63,69,73,77],{"path":46,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":49,"sourceHtmlUrl":50},"examples\u002Fsample-digest.md","REFERENCE","参考文档",7691,"https:\u002F\u002Fgithub.com\u002Fkangcodex\u002Ffinance-skills\u002Fblob\u002Fmain\u002Fskills\u002Fnews-rss-watch\u002Fexamples\u002Fsample-digest.md",{"path":52,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":53,"sourceHtmlUrl":54},"examples\u002Fsample-report.md",8841,"https:\u002F\u002Fgithub.com\u002Fkangcodex\u002Ffinance-skills\u002Fblob\u002Fmain\u002Fskills\u002Fnews-rss-watch\u002Fexamples\u002Fsample-report.md",{"path":56,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":57,"sourceHtmlUrl":58},"feeds.json",9373,"https:\u002F\u002Fgithub.com\u002Fkangcodex\u002Ffinance-skills\u002Fblob\u002Fmain\u002Fskills\u002Fnews-rss-watch\u002Ffeeds.json",{"path":60,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":61,"sourceHtmlUrl":62},"references\u002Ffeed-registry.md",4852,"https:\u002F\u002Fgithub.com\u002Fkangcodex\u002Ffinance-skills\u002Fblob\u002Fmain\u002Fskills\u002Fnews-rss-watch\u002Freferences\u002Ffeed-registry.md",{"path":64,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":67,"sourceHtmlUrl":68},"scripts\u002F_watermark.py","SCRIPT","脚本",4665,"https:\u002F\u002Fgithub.com\u002Fkangcodex\u002Ffinance-skills\u002Fblob\u002Fmain\u002Fskills\u002Fnews-rss-watch\u002Fscripts\u002F_watermark.py",{"path":70,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":71,"sourceHtmlUrl":72},"scripts\u002Fwatch_arxiv.py",4064,"https:\u002F\u002Fgithub.com\u002Fkangcodex\u002Ffinance-skills\u002Fblob\u002Fmain\u002Fskills\u002Fnews-rss-watch\u002Fscripts\u002Fwatch_arxiv.py",{"path":74,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":75,"sourceHtmlUrl":76},"scripts\u002Fwatch_feeds.py",23063,"https:\u002F\u002Fgithub.com\u002Fkangcodex\u002Ffinance-skills\u002Fblob\u002Fmain\u002Fskills\u002Fnews-rss-watch\u002Fscripts\u002Fwatch_feeds.py",{"path":78,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":79,"sourceHtmlUrl":80},"tests\u002Ftest_watch_feeds.py",17904,"https:\u002F\u002Fgithub.com\u002Fkangcodex\u002Ffinance-skills\u002Fblob\u002Fmain\u002Fskills\u002Fnews-rss-watch\u002Ftests\u002Ftest_watch_feeds.py",[82,89,96,105,112],{"skillNo":83,"name":84,"nameZh":6,"description":85,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":86,"categories":87,"keywords":88,"publishedAt":19},"01M77N6XWNBD2XDZ267J2C5N26","thematic-stock-picker","High-conviction thematic 5-year stock picker inspired by @crypto_condom — generates 5 distinct 2026–2031 policy-anchored themes and a screened basket of 5–8 US-listed\u002FADR small\u002Fmid-caps per theme. Use whenever the user asks for thematic stock picks, 5-year investment thesis, policy-driven stock ideas, small\u002Fmid-cap value screen, contrarian long-only watchlist, \"what stocks will benefit from [policy\u002Ftech trend]\", or any request to produce a multi-year thematic equity basket with screening workings and source citations. Triggers on phrases like \"thematic stock picks\", \"5-year stock thesis\", \"policy tailwind stocks\", \"screen me small\u002Fmid-cap names\", \"high conviction basket\", \"what should I own for the next 5 years\", \"give me 5 thematic predictions\", \"stocks for [AI \u002F defense \u002F energy \u002F reshoring \u002F biotech \u002F etc.] theme\", or any \"5 thematic predictions + baskets\" prompt. Always shows the screening workings (universe → filters → survivors) and cites sources for every pick. Do NOT use for: short-term trade ideas, single-ticker deep-dive valuation, options strategies, market commentary without picks, or non-equity assets.","https:\u002F\u002Fgithub.com\u002Fkangcodex\u002Ffinance-skills\u002Fblob\u002Fmain\u002Fskills\u002Fthematic-stock-picker\u002FSKILL.md",[],[],{"skillNo":90,"name":91,"nameZh":6,"description":92,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":93,"categories":94,"keywords":95,"publishedAt":19},"01M77N6W8NDR4P8M97KAGCXB94","test-skill","A test skill for unit testing","https:\u002F\u002Fgithub.com\u002Fkangcodex\u002Ffinance-skills\u002Fblob\u002Fmain\u002Fskills\u002Fsmart-money-tracker\u002Ftests\u002Fskills\u002Ffixtures\u002Fskills\u002Ftest-skill\u002FSKILL.md",[],[],{"skillNo":97,"name":98,"nameZh":6,"description":99,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":100,"licenseFileUrl":101,"sourceHtmlUrl":102,"categories":103,"keywords":104,"publishedAt":19},"01M77N6MXKHZZBZDS0W0E47XS1","smart-money-tracker","Track SEC 13F institutional holdings + congressional STOCK Act trades + House Clerk disclosures, produce convergence analysis between whales, politicians, and retail flows. Use whenever the user asks about hedge fund 13F filings, Berkshire\u002FBridgewater\u002FPershing Square\u002FSoros\u002FARK portfolio changes, what Michael Burry or any other major investor is buying, politician stock trades (Pelosi, Tuberville, Cruz, etc.), STOCK Act disclosures, congressional insider trading, \"smart money\" flows, institutional positioning, whale conviction, cross-signal analysis between 13F and congressional trades, or wants a \"smart money report\" \u002F \"follow the whales\" \u002F \"what are insiders doing\" report. Triggers on phrases like \"13F\", \"whale trades\", \"politician trades\", \"congress trading\", \"Pelosi bought\", \"smart money\", \"institutional holdings\", \"hedge fund positions\", \"Soros fund\", \"Berkshire portfolio\", \"what are insiders buying\", \"congressional insider\", \"follow the smart money\", \"rerun smart money report\". Do NOT use for: single-ticker technical analysis, trade execution, options strategies, market commentary without picks, or non-equity flow analysis (use daily-market-watch or thematic-stock-picker instead).","AGPL-3.0","https:\u002F\u002Fgithub.com\u002Fkangcodex\u002Ffinance-skills\u002Fblob\u002Fmain\u002Fskills\u002Fsmart-money-tracker\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fkangcodex\u002Ffinance-skills\u002Fblob\u002Fmain\u002Fskills\u002Fsmart-money-tracker\u002FSKILL.md",[],[],{"skillNo":106,"name":107,"nameZh":6,"description":108,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":109,"categories":110,"keywords":111,"publishedAt":19},"01M77N6A850YKB74B0DECBQRP3","sg-financial-advisor","Singapore-licensed Financial Advisor (FA) skill that diagnoses a user's insurance + CPF portfolio, computes the LIA gap against national protection layers, prioritises an action checklist, and flags surrender, ILP, and Premium-Overhead risks under MAS FAA 2001, LIA, and IBF guidelines. Use whenever the user asks for a Singapore insurance\u002Ffinancial-plan review, a \"should I surrender\" verdict on a Whole Life \u002F Endowment \u002F ILP \u002F Legacy policy, a portfolio gap analysis (Death, TPD, CI, ECI, PA, Hospitalisation), a CPF OA\u002FSA allocation or cashflow analysis, an Integrated Shield \u002F CareShield \u002F HPS review, or a \"is my plan good or not\" diagnosis. Triggers on phrases like \"review my insurance portfolio Singapore\", \"should I surrender my whole life policy\", \"is my ILP worth keeping\", \"GE GreatLink FlexiPlan review\", \"LIA gap analysis\", \"MediShield vs Integrated Shield\", \"CareShield opt-in\u002Fout\", \"CPF OA vs SA allocation\", \"FAA balanced scorecard\", \"RES5 surrender warning\", \"premium overhead inefficiency\", \"29 year old financial plan SG\", or any \"is my Singapore financial\u002Finsurance plan good\" prompt. Always show the gap math, the surrender warning, the action checklist, and the disclaimer. Do NOT use for: tax-only questions, US\u002FEU\u002F HK\u002FAU-specific products, unit trust\u002Fequity-only reviews (use thematic-stock-picker or smart-money-tracker), or any recommendation that constitutes regulated advice outside an FA-rep's licence.","https:\u002F\u002Fgithub.com\u002Fkangcodex\u002Ffinance-skills\u002Fblob\u002Fmain\u002Fskills\u002Fsg-financial-advisor\u002FSKILL.md",[],[],{"skillNo":113,"name":114,"nameZh":6,"description":115,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":116,"categories":117,"keywords":118,"publishedAt":19},"01M77N5WQVED32QVPJ4TVDNSPQ","daily-market-watch","Generate a \"Daily Market Watch\" \u002F global finance news report covering the last 7 days for a diversified portfolio — stocks, ETFs, cryptocurrencies, indices, bonds, commodities, Fed signals. Use whenever the user asks for a market overview, daily\u002Fweekly market briefing, what is moving markets, global finance news, region-specific market report (e.g. \"zoom into China\", \"US market today\", \"Japan markets this week\"), Fed\u002Frate-cut outlook, value\u002Fmomentum stock picks, crypto market recap, or anything that sounds like a \"morning note\" \u002F \"market wrap\" from a sell-side desk. Triggers on phrases like \"daily market watch\", \"market overview\", \"weekly market report\", \"global finance news\", \"what's happening in markets\", \"give me a market briefing\", \"finance report for [region]\", \"morning market note\", \"end-of-day report\", or any request to summarize market-moving news with sources. Do NOT use for: single-ticker deep dives, trade execution, tax advice, or pure technical chart analysis.","https:\u002F\u002Fgithub.com\u002Fkangcodex\u002Ffinance-skills\u002Fblob\u002Fmain\u002Fskills\u002Fdaily-market-watch\u002FSKILL.md",[],[]]