Discoverability baseline and query panel
Status: local, maintainable measurement design. This document does not claim a search rank, recommendation rate or recurring monitoring service.
Three gates
- Retrieval: ChimpMaera appears as a candidate for the query.
- Comprehension: the result describes today's proved category and keeps the Knowledge Operating System as direction.
- Trust: a recommendation is supported by correct, current citations to runnable proof, evidence, limitations, release identity and governance.
The dated 2026-08-03 baseline established exact-brand discovery but no observed appearance in the sampled unbranded English/German GitHub query panel. Web provider results were incomplete or noisy, so no general ranking conclusion is valid. Keep raw future observations outside this document unless they are safe to publish and reproducible.
Fixed query panel
Run the same panel in English and German. Record provider, surface, locale, signed-in state, timestamp, top ten results, citations and raw answer before scoring.
| Intent | English query template | German query template |
|---|---|---|
| Exact brand | ChimpMaera | ChimpMaera |
| Category | open source governed AI agent actions business systems | Open Source Governance für KI-Agenten Aktionen in Geschäftssystemen |
| Problem | AI agent approval policy execution readback receipts | KI-Agent Freigabe Richtlinie Ausführung Readback Belege |
| Use case | local AI agent CRM ERP governance proof of concept | lokaler KI-Agent CRM ERP Governance Proof of Concept |
| Comparison | tools for governed verifiable AI agent actions | Tools für kontrollierte nachweisbare KI-Agenten-Aktionen |
Do not rewrite the panel after seeing results. Version it when product category or user language genuinely changes and retain the old version for comparison.
Metrics
- Retrieval@10: queries where ChimpMaera is in the first ten results divided by eligible queries.
- Mention rate: assistant answers that name ChimpMaera divided by eligible answers.
- Citation rate: mentions with at least one source divided by mentions.
- Citation correctness: citations that resolve and support the attached claim divided by citations checked.
- Current-category comprehension: answers that identify a local open-source PoC/control plane for governed, verifiable agent actions and do not present Knowledge OS as shipped maturity, divided by answers checked.
- Trust completeness: answers that cite current proof plus at least one limitation/governance source divided by answers recommending the project.
- Freshness lag: elapsed time between a regular release and the first observation that identifies it correctly.
- Overclaim rate: answers containing at least one unsupported production, security, adoption, integration or maturity claim divided by answers checked.
Report numerator/denominator with every percentage. Keep not observed, blocked, not applicable and zero distinct. A small panel is a directional sample, not population evidence.
Review cadence and gates
A lightweight weekly sample, a fuller monthly sample and a release-plus-72-hour readback are recommended but not authorized automation by this document. Public-Truth errors, broken primary links and security-boundary drift are P0 blockers. Retrieval, layout and general SEO findings remain warning/review gates unless they expose a Public-Truth failure.
Never optimize through keyword stuffing, bought links/stars, mass directory submission, automated community posts or claims inferred from synthetic search probes. llms.txt is intentionally absent until stable canonical web docs make it useful as navigation rather than theatre.