research

Research skill

Three prompts the pipeline uses to convert raw signal into topic decisions. The architect routes on the description above; UnifiedPromptManager resolves each template by key (Langfuse override still wins over the bodies below).

Default prompts — basic but functional; production-quality prompt packs ship as a premium add-on.

research.analyze_search_results

Analyze these search results and return JSON with keys: summary, key_points (list), sources (list).
Search results: {search_results}

topic.ranking

You are scoring topic candidates for a content pipeline against the operator's weighted goals.

Goals (weight in pct):
{weights_descr}

Candidates:
{cand_block}

Return STRICT JSON keyed by candidate id, mapping each id directly to its score
as a number from 0 to 100:
{{"<id>": <score 0-100>, ...}}

Return ONLY the JSON, no commentary.

Why score-only (Glad-Labs/poindexter#926). This prompt used to also ask for a nested breakdown object keyed by "<GOAL_TYPE>". Asking the model to invent those key names was the single biggest failure surface: Langfuse traces show 10 of 19 real calls over 14 days failing to parse, and the short failures die inside a degenerate repetition loop on a breakdown key — "BRAND own own own own …", "EDU//////////////…". Even calls that did parse carried invented keys ("BREAKDOWN", "BRAND own", "BREAKING_NEWS"). Every failure fell back to the raw embedding pre-rank for the whole batch.

The breakdown was never worth generating: topic_ranking.weighted_cosine_score already computes a real per-goal breakdown from the goal vectors (plus a _grounding observability key), stores it on the candidate — and the LLM's hallucinated version then overwrote it. Asking only for the score removes the failure surface and keeps the better, calculated data (feedback_machine_rules: prefer calculated over generated).

research.distill_topic_angle

Read the snippets from an AI-operated content business's internal records.
You are looking for a STORY worth telling the audience below — something that changed or broke, a decision that was made, or a lesson that was learned. Routine operational status (clean cycles, monitoring chatter, internal tooling housekeeping) is not a story.

Audience / niche: {niche_context}

Snippets:
{joined}

If the snippets contain a story, return STRICT JSON: {{"topic": "<short title>", "angle": "<one-sentence framing: why this matters / what we learned>"}}.
If they do not, return STRICT JSON: {{"storyworthy": false, "reason": "<one short phrase>"}}.