Prepare optional source-grounded public historical context for later memoir questions, independently after the visible reply is saved.
Resources
4Install
npx skillscat add bohui/copyme2-ai/memoir-historical-news-research Install via the SkillsCat registry.
SKILL.md
Historical news research
This is an application-owned background lane, not a collector tool. The collector
consumes only an already-ready projection. Never run research during its reply,
add a classifier, infer biography from a public event, or change photo defaults.
Use the deterministic adapter in apps/api/historical_news.py and gateway adapter
in apps/api/historical_news_provider.py. Their input is accepted versioned life
context; their only output is external contextual clues. See the contract below.
- Favor the active personal memory and ordinary retained questions first.
- Missing, conflicting, excluded or overly broad context means no lookup.
- Birthplace is tentative geography, never proof of later residence.
- Discovery uses only reviewed public place labels, coarse years, neutral terms.
- Require a completed native search receipt, returned URL and inspected page.
- Public event dates are distinct from source publication and personal dates.
- A shown clue produces no narrator evidence or MemoryEvent. A later answer is
the storyteller's exact testimony, with the cue separately attributed. - At most one ready clue joins the five-candidate pool. Exactly one question is
visible, in the current interview language, with the source title preserved. - Invite recall neutrally; no assumed awareness, attendance, residence or emotion.
- Permit skip and uncertainty. Retire shown/rejected clues; opt-out fences jobs.
- Never select unsolicited violence, trauma, discrimination or political conflict.
- No credentials, paid plans, production migrations or live evaluations are
authorized by this skill. Default rollout is off.
Read references/contract.md for budgets, evidence, privacy and review boundaries.