The Kardec skill’s creator claims the released package grounds agent answers in seven primary Spiritist texts, potentially reducing fabricated quotations and citation numbers in that domain.
state: seedheat: lowuncertainty: mediumknownscott: lowsource-grounded-rag agent-skills citation-provenanceSideSuspicious8083iaespirita
What is this?
A creator identified in the case as SideSuspicious8083/iaespirita says they released an agent skill that packages seven Allan Kardec works to ground answers and reduce invented quotations or citation numbers. The web results establish that Kardec’s writings are primary texts of Spiritism and that its conventional “basic works” comprise five books, but they do not surface the skill, its repository, its seven-work composition, implementation, or any evaluation of citation accuracy. The package’s release and effectiveness therefore remain creator claims supported here only by the supplied evidence titles.
Why it matters to Scott
Scott already holds the core position in “Witness Not Oracle,” “Evidence Package,” and his proof-of-read citation gate: generated claims should carry verbatim, source-resolvable evidence rather than model-invented references. This is an unvalidated niche implementation of that established pattern; without implementation details or citation-accuracy evaluation, it does not extend or challenge his work.
ip:source.witness-not-oracle-ebookip:concept.evidence-packagedev:concept.proof-of-read-citation-gatedev:concept.verbatim-source-evidence-anchoringradar:concept.agent-skillsradar:concept.citation-verification
queries asked of Scott's wikis
- source-grounded RAG citation verification
- agent skills packaging domain corpora
- quotation provenance and citation hallucinations
- primary-source retrieval versus model memory
- RAG evaluation for exact quotes and references
- maintaining curated knowledge packs for agents
Measured heat
now 0 pts/hpeak 0 pts/hcomments 0/hpeers p0momentum: steady2 platformsage 519h
points/hour across evidence · reading as of 2026-10-12 02:59:37.977291+11:00 · deterministic, not a model opinion
How the heat travelled
pace: p44 vs 1032 stories at the 336h mark (now 519h old) — ahead of anthropic-meta-lawsuit (1.2x), behind legion-elixir-lua-agent-sandbox (0.9x)
Evidence (2) — ⭐ canonical anchor
Interpretation history
2026-09-20T02:26:29Z
grounded: known/low — Scott already holds the core position in “Witness Not Oracle,” “Evidence Package,” and his proof-of-read citation gate: generated claims should carry verbatim,
2026-09-20T02:23:25Z
origin walked (codex/luna, conf 0.98): anchor reddit.post.1wl3ehp -> echo.github.0992a8ba46 by Bruno Winícius Amorim (GitHub: bwasistemas)
2026-09-20T02:21:40Z
case created — The installable package is a concrete source-grounding artifact, but its citation-reliability benefit is supported only by the creator’s low-engagement announcement.
Decision trace
- 09-25 12:26review_screenjev screen: no material development (noul=0.06)
- 09-20 12:26groundScott already holds the core position in “Witness Not Oracle,” “Evidence Package,” and his proof-of-read citation gate: generated claims should carry verbatim, source-resolvable evidence rather than m
- 09-20 12:23promote_anchororigin walk conf 0.98
- 09-20 12:21createThe installable package is a concrete source-grounding artifact, but its citation-reliability benefit is supported only by the creator’s low-engagement announcement.