HCS-25 (Appendix): Trust Signal Catalog (Informative)

This document complements HCS-25 by describing trust signal concepts and providing pointers to the per-signal catalog.

It is informative-only: the exact signals and providers are expected to vary by ecosystem.

Prefer the per-signal catalog under ./signals/index.md for the current, maintainable structure.

Quick links:

• Signal catalog (index)
• SimpleMath / SimpleScience eval methodology (signal family)

Signal record shape (recommended)

A signal record SHOULD be represented as a small object that includes:

• status: ok | missing | timeout | error | stale
• fetchedAt: an ISO timestamp indicating when the signal was last refreshed
• sources: a list of stable source identifiers (strings) describing where the signal was derived from
• data: a provider-neutral payload containing the values needed by scoring adapters

Example (illustrative):

{
  "status": "ok",
  "fetchedAt": "2026-01-14T00:00:00Z",
  "sources": ["example:monitoring:v1"],
  "data": {
    "uptime_30d": 0.9981,
    "p50_latency_ms": 210
  }
}

Signal identifier namespace pattern (normative)

Conforming implementations MUST use stable, namespaced signal identifiers. HCS-25 defines a dot-separated namespace convention; see Signal Identifier Namespacing.

In short:

{namespace}.{name}[.{subname}...]

Where each segment is lowercase ASCII and matches /[a-z][a-z0-9_-]*/.

Implementations MUST document their chosen signal IDs and how they map to adapters.

Per-signal catalog (informative)

The maintainable catalog lives under ./signals/index.md and includes per-signal schema pages (examples) for:

• marketplace ecosystems (ACP, AgentVerse, ERC-8004),
• operational signals (availability),
• adoption signals (OSS popularity),
• usage signals (x402), and
• model eval signals (OpenRouter, Chatbot Arena, Hugging Face, Open LLM Leaderboard).

For adapter mappings, see Adapter Catalog or ./adapters/index.md.

Stored in: subject metadata.additional.

Field	Type	Meaning
chatbotArenaEvalScore	number | null	Normalized score [0,100] computed from min/max Elo observed in the snapshot
chatbotArenaEvalElo	number | null	Raw Elo-like value from the source table
chatbotArenaEvalVotes	number | null	Vote/sample size indicator (if present)
chatbotArenaEvalStatus	ok | missing	Availability status
chatbotArenaEvalUpdatedAt	ISO timestamp	Refresh time
chatbotArenaEvalSources	string[]	Sources list (e.g., chatbot-arena:openlm)

huggingface-model-index (model metadata + eval-like aggregation)

Stored in: subject metadata.additional.

Field	Type	Meaning
huggingFaceModelId	string | null	Canonical HF model id
openrouterHuggingFaceId	string | null	HF id resolved via OpenRouter model list
huggingFaceEvalScore	number | null	Normalized score [0,100] derived from model-index metrics and/or popularity
huggingFaceEvalMetricsCount	number	Count of included metrics
huggingFaceDownloads	number | null	Popularity proxy
huggingFaceLikes	number | null	Popularity proxy
huggingFaceEvalMode	model-index | popularity | mixed | missing	How the score was produced
huggingFaceEvalStatus	ok | missing | error	Availability status
huggingFaceEvalUpdatedAt	ISO timestamp	Refresh time
huggingFaceEvalSources	string[]	Sources list (e.g., huggingface:model-index, huggingface:popularity)

openllm-leaderboard (Open LLM Leaderboard dataset)

Stored in: subject metadata.additional.

Field	Type	Meaning
openLlmEvalScore	number | null	Normalized score [0,100] derived from selected benchmark metrics
openLlmEvalMetricsCount	number	Count of included metrics
openLlmEvalStatus	ok | missing | error	Availability status
openLlmEvalUpdatedAt	ISO timestamp	Refresh time
openLlmEvalSources	string[]	Sources list (e.g., openllm:leaderboard:v2)