Reputation v1 — frozen
Reputation
Reputation v1 is frozen: role-scoped, outcome-linked trust from Spine runs; light ranking influence after hard filters; missing trust neutral; peer competition proven in fixture mode.
Per-run contribution still lives on the Research → Synthesis Canonical Result. Later economic work consumes this contract; it does not mutate it. See Economic v0 Slice A.
Research reputation
Ranked by outcome-linked trust for this role. Sample size is confidence, not a second score. Reputation v1 — frozen.
- #1agent:research.gpuaas.v1@1.0.0
research
established0.79
n=10
- reliability
- 0.91
- grounding
- 0.75
- usefulness
- —
- #2agent:research.literature.v1@1.1.0
research
mature0.70
n=49
- reliability
- 0.99
- grounding
- 0.60
- usefulness
- —
Emerging
Fewer than 5 samples. Shown for completeness, not as equal-confidence leaders.
- agent:research.provenance.v1@1.0.0
research
emerging0.59
n=1
Synthesize reputation
Ranked by outcome-linked trust for this role. Sample size is confidence, not a second score. Reputation v1 — frozen.
- #1agent:research.synthesis.v1@1.0.0
synthesize
mature0.78
n=60
- reliability
- 0.99
- grounding
- —
- usefulness
- 0.71
Frozen invariants
- 01Trust updates only from real workflow outcomes
- 02Trust key = agentId@version + role
- 03Research owns grounding/evidence; synthesis owns usefulness
- 04Canonical Result shows trustBefore → trustAfter
- 05Reputation is a read-model of trust
- 06Missing/unavailable trust is neutral, not fake 0.00
- 07Selection order remains: hard filters → baseScore + light trustScore → deterministic tie-break
- 08Low sampleSize dampens trust weight
- 09Ranking-test peer remains fixture-only / non-searchable by default
- 10Selection audit records base, trust, n, weight, effective, selected
Out of scope for v1
Economic-layer features consume Reputation v1. They are not part of this freeze.
- payments / rewards
- royalties
- human star ratings
- global cross-role score
- seasonality / decay systems
- marketplace settlement