laohan-redian

①Topic decision

Hot topics are only used for discovery; the final result is an auditable production decision: why do it now, who is it for, what makes the argument valid, which duplicate angles exist on Douyin, why the other candidates were eliminated, and how to review after publishing according to a fixed window.

Boundaries

• This skill is the sole lead writer of ①, writing signals, candidates, source-health, and 00-选题.*。
• laohan-douyinsousuo only writes 00-抖音搜索证据.{json,md}; the same Agent host executes according to bianpai routing, and does not make programmatic skill-to-skill calls.
• ② only writes the talking-head script; ① does not generate body text, outline paragraphs, or cover.
• ⑤ only performs complete fact-checking; ① must still find at least one PRIMARY original source for the final factual premise.
• Do not install new browsers, crawlers, Python runtime, or third-party skills. For Douyin platform access, use only the installed OpenCLI.
• The same event can form an opinion candidate and a tutorial candidate; when the audience task, title promise, and content path are truly different, they can enter comparison separately. ②—④ follow the selected lane and do not mix writing.

Input

Episode mode must read:

1. This episode's episode-config.json and 00-编排/ frozen artifacts.
2. Feedback snapshots explicitly registered for this episode; if none, record NOT_AVAILABLE, and do not scan old episodes to guess experience.
3. This round's real signals and Douyin search evidence.

The standard Episode is fixed as REVIEW_GATED: AI independently completes discovery and candidates, but must not choose on Jeffrey's behalf. When Jeffrey explicitly gives a topic, go to USER_SEED, and still must provide at least one real alternative candidate, Douyin evidence collection, PRIMARY evidence, and a measurement contract; the final result also goes through emotional and expressive-desire screening.

When what Jeffrey gives is a software, method, or series direction, first go toUSER_DIRECTION_RESEARCH: read the projectdocs/系列选题研究与分期规格.md, and complete targeted searches on four platforms, content-level breakdown, demand judgment, series phasing, and single-episode material package. It does not run unrelated 9-account broad scans to pad hot topics. OnlyDEMAND_VALIDATEDcan enter a series of three episodes or more,EXPERIMENTAL_ONE_OFFenters only a single episode,DIRECTION_REJECTEDstops. A formal episode must still provide alternative candidates and go through Jeffrey's emotional screening.

Workflow

1. Collect current signals

Write first00-选题-signals.json.discovery_mode: no direction isAUTONOMOUS_SCAN; Jeffrey has already given direction asUSER_DIRECTION_RESEARCH. The latter writes thenode scripts/check-series-research.mjsthat has passedseries_research_pathand the SHA intodirection_research, and registers the four routes of Bilibili, Xiaohongshu, Douyin, and Zhihu separately as signal sources.DEMAND_VALIDATEDAt least two platforms have real demand signals;EXPERIMENTAL_ONE_OFFenter a single episode only based on verified official new events, Jeffrey's personal experience, or niche-task evidence in the research package, and do not force fabrication of popularity on two platforms.

Episode mode executes the sole deterministic entry point:

node ~/Documents/laohan-skills/laohan-redian/scripts/collect-topic-signals.mjs \
  --episode episodes/<slug>

AUTONOMOUS_SCANThe candidate sources are fixed as three lines: recent scanning of 9 registered Douyin benchmark accounts account by account, comprehensive hot-topic scanning,script-pool/Jeffrey个人表达池.md. The default discovery and display priority isBENCHMARK_CREATOR(1) > BROAD_HOTSPOT(2) > PERSONAL_EXPRESSION(3); this is the topic-finding efficiency order, not an automatic selection score; personal topics can still win after obtaining public interest evidence. The default routes for comprehensive hot topics are AIHOT, Hacker News, Zhihu, Weibo, 36kr, Bilibili, Douyin Hot List, Toutiao, Tieba, and Hupu; if one route fails, record it truthfully, and the other hot topic routes continue, but if any of the 9 benchmark accounts fails, candidate generation stops, and a normal empty result is recordedEMPTYand counted as complete. Only pass this when the hot topic route needs to be narrowed --sources, and this cannot be used to skip the benchmark accounts and the personal pool.

The script only writes:

• 00-选题-signals.json
• Initial 00-选题-source-health.json

The script does not select topics, does not access old episodes, and does not call douyin-ai.js。

2. Standardize, deduplicate, and form candidates

Read signals and cluster the same factual event by stable signal id; exclude duplicate expressions with the same URL/promise/path. Benchmark accounts are only used to distill motifs, public needs, and anomalous signals, not to copy titles, copy, cases, or conclusions. The same event generates two forms only when "help the audience make a judgment" and "take the audience through completing a task" each hold separately. Write schema 3 00-选题-candidates.json, first use the top-level screening_summary to leave the filtering chain, then write at least two genuinely different shortlist candidates.screening_summary contains only:

• raw_signal_count: the actual total number of results in signals;event_cluster_count: the total number of event clusters after deduplication.
• longlist: when there are enough valid events, record 8–12 items; when there are fewer than 8, record all of them. Each item contains a unique event_cluster_id、working_title, a non-empty signal_ids, a positive integer source_count、disposition: SHORTLISTED|REJECTED , and a specific reason。
• Each final candidate's event_cluster_id must be marked in the longlist SHORTLISTED. Do not add intermediate files, extra nodes, or a second scoring system.

For the same event, viewpoint/tutorial variants may be counted separately when the audience task, title promise, and content path differ. Each item must have:

Before forming candidates, restore the original analysis: mark relevance using direct terms such as "AI/GPT/Claude/large model/robot/automation/programming/side hustle" and indirect terms such as "education/career/consumption/entrepreneurship/workplace", but retain marginal signals that may generate new angles; count content types such as installation tutorials/advanced tips/methodology/comparative reviews/hands-on/resource recommendations; mark multi-platform resonance, single-platform exclusivity, crowded angles, and opportunities not covered by competitors. They are used to expand candidates, not to replace the later PRIMARY and Douyin keyword verification.

• Unique id, stable non-empty event_cluster_id, non-empty why_now、title、content_form: opinion-video|tutorial-video、audience_level: BEGINNER|INTERMEDIATE|ADVANCED、audience_problem、audience_promise、thesis。
• candidate_origin must register primary_lane, the corresponding priority_rank, non-empty origin_signal_ids and corroborating_lanes. Cross-verification can improve credibility, but it cannot disguise a personal expression source as a benchmark viral source.
• tension must register type: CONFLICT|MISCONCEPTION|COUNTERINTUITIVE|TRADEOFF、common_assumption、jeffrey_position and a conflict_statementthat can be fulfilled by the main text in one sentence. Writing only a domain name, product launch, or "worth paying attention to" does not count as a candidate.
• creator_fit must clearly state basis: FIRSTHAND_EXPERIENCE|BUILT_OR_TESTED|OWNED_EVIDENCE|LONG_TERM_PRACTICE|RESEARCHED_JUDGMENT|NOT_ESTABLISHED、why_jeffrey and distinctive_judgment. It answers "why Jeffrey should talk about it, what his basis is, and what judgment it adds compared with a generic retelling." SELECTED does not allow NOT_ESTABLISHED; when there is no firsthand experience, RESEARCHED_JUDGMENTmay be used, but it must form an independent judgment with evidentiary boundaries, and "I also paid attention to it" cannot be passed off as a personal advantage.
• assumed_prerequisites and jargon_to_explain are string arrays. BEGINNER candidates must include the key tools, commands, and migration/import concepts in the title and argument in the explanation checklist, and cannot merely increase the audience_fit score.
• Non-empty signal_ids and evidence_ids。
• Non-empty platform_query_intent；platform_alignment contains status: ALIGNED|ADJACENT|CONTRADICTED|UNRESOLVED|UNAVAILABLE, non-empty rationale and evidence_ids. When there are platform results, evidence ids must be real; when platform access fails, write UNAVAILABLE and an empty array.
• content_gap contains status: CONFIRMED|NOT_CONFIRMED|UNAVAILABLE, non-empty rationale and evidence_ids. Only when the platform directly provides search volume, supply or gap evidence can the first two be written; search result counts, missing interaction fields, or model judgments must never be passed off as content gaps.UNAVAILABLE is legal and does not participate in mechanical elimination.
• non-empty claim_evidence_map; each entry contains a unique claim_id, non-empty claim、evidence_ids. It must bind technical facts to PRIMARY; when platform evidence is available, bind "what people on the platform care about" to PLATFORM_SIGNAL, and must not let an adjacent keyword endorse another argument. When the platform UNAVAILABLE , do not fabricate a PLATFORM_SIGNAL claim.
• scorecard seven-dimension integer 1—5:audience_fit、evidence_strength、platform_relevance、differentiation、creator_fit、production_feasibility、learning_value。
• non-empty rationale. Before Jeffrey's screening, all use disposition: AWAITING_JEFFREY; AI only displays candidates, evidence and risks, and must not directly mark SELECTED.

Scores are challengeable candidate comparisons, not traffic forecasts. Heat cannot replace audience value, evidence strength, differentiation or filmability. SELECTED can be ALIGNED, ADJACENTwith a clear differentiation explanation, or, when discovery and PRIMARY are sufficient, UNAVAILABLEthat honestly explains the scope of the gap; platform absence lowers platform_relevance confidence, but cannot eliminate a candidate on its own.CONTRADICTED|UNRESOLVED must not be selected. Before a tutorial candidate is written as definitive steps, it must still be registered in tutorial_proof , within this period status: VERIFIED, non-empty evidence_path、evidence_sha256and version_boundary . recovery。

3. Perform Douyin keyword verification on the shortlist

Design a specific query for each of 1—3 shortlisted candidates, and have the same host execute laohan-douyinsousuo episode mode. Must obtain 00-抖音搜索证据.json and .md before continuing.

• OK: cite real result ids, and clearly state the audience intent that the query actually proves. When it directly supports the candidate title promise, audience task and why_now , mark ALIGNED; when it only supports the same product, migration heat or an adjacent concept, mark ADJACENT. Prefer adding more direct queries, but do not automatically eliminate a differentiated new angle just because it lacks content with the exact same title.
• EMPTY_OR_FIELD_UNAVAILABLE: write only "platform evidence pending", and must not write "no one is doing it/blue ocean".
• OpenCLI login or adapter all failed: keep the real FAILED evidence, mark platform_alignment and content_gap as UNAVAILABLE; as long as discovery and PRIMARY are sufficient, continue comparing, and do not treat platform absence as a blocker.

4. Return to the PRIMARY original source

For the final candidates' factual premises, look up official websites, official announcements, papers, product documentation or original public records. Platform hot lists, search results, and media paraphrases can only be PLATFORM_SIGNAL or SECONDARY。

Write PRIMARY and PLATFORM_SIGNAL evidence into the candidate file; each entry has a stable id、source、source_type、url、retrieved_at. When PRIMARY is missing, narrow the argument to the provable scope; if it still cannot hold, switch candidates or BLOCKED.

5. Jeffrey emotion and expression desire screening

After the candidates are complete, enterWAITING_FOR_JEFFREY_EMOTION_SELECTION. Display 3—5 candidates at a time, and ask Jeffrey to choose one based on a genuine first reaction, or explicitly choose "all feel indifferent".

• Select one: recordfirst_reaction、challenge_or_addition、firsthand_detail、would_say_without_heat=true. First change the sole candidate toSELECTED, and the rest toREJECTED, then write schema 1 00-选题-Jeffrey筛选.jsonBind the current candidate SHA and the unique candidate ID.
• All no-preference: writestatus=REJECTED_ALL, bind the current candidate SHA, and recordrejection_reason、strongest_near_missandrescan_direction. EnterTOPIC_RESCAN_REQUIRED, and in the same 01 task return to the three-way scan; do not force SELECTED, generate a final topic, or enter ②. Before a new round of candidates replaces the old candidates, these four items must be written as-is along with the old candidate SHA into the new00-选题-candidates.json.screening_summary.rejection_history[]; validator uses this to switch fromTOPIC_RESCAN_REQUIREDback toWAITING_FOR_JEFFREY_EMOTION_SELECTION. The final acceptance record may overwrite the screening file, but the rejection history remains in the candidate file.

Without a SHA-closed Jeffrey record, no final topic may be generated.

6. Write the final decision and source health

Update 00-选题-source-health.json, recording for each channel:

• Unique source_id and source_role: DISCOVERY|BENCHMARK_CREATOR|PERSONAL_EXPRESSION|DOUYIN_SEARCH|PRIMARY_PROOF. schema 3 must retain the latter two local discovery roles; the personal expression pool allowsEMPTY, while benchmark accounts only allow the fully scannedOK|EMPTY。
• Actual command_or_url、attempted_at、status、result_count。
• For this period relative to result_file and actual result_sha256。
• FAILED writes error; SKIPPED writes reason。

The status must be consistent: the number of OK entries is greater than 0, and the number of EMPTY/FAILED/SKIPPED entries is 0. At least one DISCOVERY is OK, one DOUYIN_SEARCH has a real attempt record, and one PRIMARY_PROOF is OK; DOUYIN_SEARCH may be FAILED, but must be bound to this period's Douyin evidence file and the original error.

Write schema 3 00-选题.json, retain the original schema 2 fields, and bind the candidate selected by Jeffrey. Finally execute:

node ~/Documents/laohan-skills/laohan-redian/scripts/check-topic-contract.mjs --episode episodes/<slug>

Only outputPASS redian topic contract schema=3counts as ① complete.

• schema_version: 2、selected_candidate_id, non-empty rejected_candidate_ids。
• Copy as-is the selected item's content_form、audience_level、audience_promise、creator_fit、assumed_prerequisites、jargon_to_explain、platform_query_intent、platform_alignment、content_gap and claim_evidence_map。
• audience、thesis、evidence、selection_rationale、not_do_reason。
• All evidence belongs to the selected candidate, and at least one is PRIMARY; when the platform is available, at least one PLATFORM_SIGNAL,platform_alignment.status=UNAVAILABLE may be absent when .
• experiment contains a unique primary intervention, legal metric_keys, positive integer T+N、observation_window_unit: DAY and per-key metric_targets。

Each metric target:

{
  "key": "avg_duration_sec",
  "direction": "INCREASE",
  "baseline_ref": "recent_5_median",
  "measurement_source": "douyin_creator_center"
}

Finally write 00-选题.md, explaining only the real decisions in the machine contract, without adding unregistered facts.

Failure handling

Trigger condition	Handling
A single discovery source fails	Record FAILED; other sources that satisfy the role contract may continue
All discovery sources have no results	BLOCKED, do not choose a topic from model memory
Douyin returns a legal empty array	Record EMPTY; do not infer that there is no competition
OpenCLI login/adapter all-path failure	Record platform evidence as UNAVAILABLE and continue; do not install an alternative scraper, do not infer a blue ocean
The final candidate has no PRIMARY	Narrow the thesis or switch candidates; if still none, BLOCKED
Jeffrey has no preference for any candidate	WriteREJECTED_ALLand enterTOPIC_RESCAN_REQUIRED; rescan according to the rejection reasons, do not enter ②
Platform results support only adjacent topics	Mark ADJACENT, supplement the query; if still adjacent, truthfully explain the differentiation and risks, may participate in the final comparison, must not pretend to be directly aligned
The tutorial candidate has no real local execution evidence	Keep REJECTED; may be retained as a pending-verification candidate, must not be written as a tutorial that is already deliverable
Candidate cannot be measured under the existing ⑬ metric	Switch intervention/metric or switch candidate

Anti-patterns

❌ “All three platforms are hot, so let’s do this one.”

✅ “Candidate C02 simultaneously addresses the target audience problem, PRIMARY supports the factual premise, and Douyin evidence shows existing content is concentrated on feature introductions, while this issue differentiates through the cost of real workflows; C01/C03 were eliminated due to weak evidence or being unfilmable.”

❌ “Some people on Douyin searched for Claude Code migration, so that proves the ‘validation loop’ is a current platform demand.”

✅ “That result only proves migration/import demand; mark it ADJACENT; continue searching for direct intent such as validation failures, launch regressions, etc. If there are still no direct results, you may include the ‘validation loop’ as a differentiated viewpoint in the comparison, but make it clear that platform evidence only supports an adjacent demand.”

❌ “0 search results, so this is a blue ocean.”

✅ “The search executed successfully but returned 0 results; record this as platform evidence undetermined, and do not use it to judge competition.”

Non-Episode mode

When the user only wants a hot-topic brief, you may run the same collection script into a temporary directory and then return a summary of the current signals; do not write to the episode source of truth. When the user only wants to search Douyin, route directly laohan-douyinsousuo。