Menu
Detecting and filing knowledge-base gaps from failed groundings, conflicts, staleness, and ticket-theme analysis; drafting KB stubs with evidence. Owned by Metis; consumed by Echo.
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Baseado na classificação ocupacional SOC
Offline A/B experimentation for prompt/policy/skill/rubric variants: defines what constitutes an experiment, the offline-only rule, the scoring contract, the anti-Goodhart + safety-floor constraints, the statistical-honesty rule, and the governance wiring into the TheEights evolution loop.
Encoded policy constraints: refunds/disclosures/consent, GDPR/CCPA data-minimization themes, right-to-human mandates, regulated-claim language rules, AI-disclosure requirements. Consumed by Harmonia, Soteria, Hermes, Eunomia.
The structured state token that travels with the customer across every agent handoff — identity ref, goal, active objects, constraints, sentiment, history digest. Minted by Iris; carried inside HANDOFF payloads.
Hard budgets, bounded loops, and the four terminal states. Step/dispatch/retry ceilings per command, graph-level termination conditions, and cost-tiering rules. Owned by Hestia.
The KPI frame: containment, FCR, AHT, CSAT-proxy, escalation precision/recall, cost-per-resolution, SLA attainment — and the Agent-Manager dashboard contract over events.jsonl. Owned by Hestia; consumed by Themis, Soteria, Echo.
Intent/language/sentiment/priority classification and routing rules for incoming support tickets. Owned by Iris; consumed by Hestia for dispatch.
| name | kb-gap-detection |
| description | Detecting and filing knowledge-base gaps from failed groundings, conflicts, staleness, and ticket-theme analysis; drafting KB stubs with evidence. Owned by Metis; consumed by Echo. |
| user-invocable | false |
| argument-hint | <file-gap|draft-stub|gap-report> |
| allowed-tools | ["Read","Grep","Glob"] |
Every NO_ANSWER_SAFE_FALLBACK is the KB telling the house where it is thin. This skill turns grounding failures into a curation pipeline: the squad does not edit the KB, it files evidence the KB team can act on.
Detect gaps systematically, file them with enough evidence to fix, and draft candidate KB stubs from resolved tickets — the "Knowledge Management Agent" function from the research taxonomy.
kb-rag-citation)./kb-gap-report for the periodic roll-up.| Class | Signal |
|---|---|
| Missing | retrieval returned nothing usable for a real intent |
| Stale | volatile-class source disqualified by age (freshness-aware-retrieval) |
| Conflicting | two sources disagree on currently-true facts |
| Unfindable | the content exists but retrieval missed it (vocabulary mismatch between customer language and doc language) |
| Wrong-altitude | doc exists but answers the adjacent question (reference where how-to is needed, or vice versa) |
Filed to hearth/output/kb-gaps/{topic}-{date}.md:
kb_gap:
gap_id: GAP-<date>-<seq>
class: missing | stale | conflicting | unfindable | wrong-altitude
intent: <triage intent class>
customer_phrasing: [<redacted phrasings customers actually used>]
tickets: [<ticket ids>] # evidence trail
sources_involved: [<doc refs + as-of dates>] # for stale/conflicting
frequency: <count this period>
proposed_fix: <new doc | update + which | retire + which | retitle/alias>
For missing gaps with a RESOLVED ticket trail, draft a candidate stub
from the resolution: problem statement in customer phrasing, the verified
steps (from Asclepius's diagnosis or Metis's eventual grounded answer),
applicability bounds, and an explicit DRAFT — needs owner review header.
Stubs are candidates for the KB team, never silently published; the
squad's authority ends at evidence.
The customer_phrasing field is the quiet gold: docs written in product
vocabulary fail retrieval against customer vocabulary. Echo aggregates
phrasings per theme so fixes get titled in the words customers will
search with.