Each pack answers a distinct phase question in the care cycle. The measures below are assigned so that each metric belongs to exactly one pack.
These metrics are designed for sufficiency, not maximisation. Each deployment context defines a threshold — "good enough" for that community. Crossing the threshold is the goal; score-chasing beyond it risks the same metric-gaming the 6-Pack warns against. The kami tends its garden — it does not compete to tend the most gardens.
Pack 1 — Attentiveness: did we look at the right things?
| Metric | What it answers |
|---|
| Coverage | What share of affected people provided input? |
| Absence rate | Which demographic segments have no voice in the bridging map? |
| Voice equity | How much facilitated time do least-heard groups get vs. most-heard? |
| Bridging map | Which stakeholder clusters are represented, and which cleavages are reinforcing vs. cross-cutting? |
Pack 2 — Responsibility: did we make and keep the right promises?
| Metric | What it answers |
|---|
| Promise coverage | What share of identified needs have a named owner and SLA? |
| SLA adherence | What share of cases are resolved within their published response window? |
| Adopt-or-explain rate | What share of Assembly outcomes have documented adoption or published deviation with remedy? |
Pack 3 — Competence: did we execute correctly?
| Metric | What it answers |
|---|
| Decision accuracy | What share of decisions are overturned on audit or appeal? |
| Guardrail integrity | What share of red-line tests pass? |
| Trace completeness | What share of decisions have a full trace (rule, source, uncertainty, receipt)? |
| Canary health | What share of canary releases complete without triggering rollback? |
Pack 4 — Responsiveness: did the care land well?
| Metric | What it answers |
|---|
| Resolution rate | What share of pause-triggering cases are resolved within the promised window? |
| Appeal closure time | How does actual vs. target appeal closure time compare across case types? |
| Harm recurrence | At what rate do resolved incidents re-appear within 90 days? |
| Trust-under-loss | What is the trust score after a bad outcome — did repair work? |
Pack 5 — Solidarity: is the ecosystem structurally fair?
| Metric | What it answers |
|---|
| Bridge index | What are the cross-group participation and endorsement rates in shared decisions, published quarterly? |
| Portability rate | What share of users successfully export data when leaving? |
| Agent ID coverage | What share of agents have verifiable meronymous attestations? |
Pack 6 — Symbiosis: is the system bounded and sustainable?
| Metric | What it answers |
|---|
| Scope compliance | What share of agent actions fall within declared purpose bounds? |
| Succession readiness | When was the last successful exit drill? |
| Sunset compliance | What share of agents have current attestation and active sunset timer? |
| Ecology diversity | How many independent agents serve equivalent needs in the same domain? |
Trust decomposition
Trust is present across all six phases, but each pack measures a distinct dimension:
| Pack | Trust dimension |
|---|
| 1 — Attentiveness | Trust in being heard (voice equity) |
| 2 — Responsibility | Trust in promises (SLA adherence, adopt-or-explain) |
| 3 — Competence | Trust in execution (trace completeness, guardrail integrity) |
| 4 — Responsiveness | Trust after harm (trust-under-loss) |
| 5 — Solidarity | Trust across groups (bridge index) |
| 6 — Symbiosis | Trust over time (succession readiness, scope compliance) |
Cross-group endorsement appears in two distinct roles: as an RLCF training signal in Pack 4 (a training objective — how you shape the model) and as the basis of the bridge index in Pack 5 (an ecosystem audit — what you report). These are separate uses.
