Record grant lifecycle live proof

This commit is contained in:
Meta Vens 2026-08-11 08:13:04 +02:00
commit 56cd8ef375
4 changed files with 109 additions and 18 deletions

View file

@ -3853,3 +3853,36 @@ Directory observation is a bounded source faculty, not Heartbeat authority.
tool in copied live cognition and measured 72 directories, 74 files, and
3,975,362 bytes with `complete=true`. Modeling refused to infer post-brake
grant terminality from artifact names, preserving the ownership boundary.
## Grant-owned resident lifecycle sight
Resident lifecycle observation and scheduling now read the same authority.
- Owner: each exact `ResidentSelfHeartbeatGrant` owns its consumed and terminal
state; `ResidentSelfState` owns whether that grant is prepared or active.
- Inputs: one explicitly bound resident store, an optional exact grant ID, and
a bounded recent-row limit.
- Outputs: typed pressure/schedule/action identity, consumed and terminal
timestamps/status, active/prepared flags, and derived `launchable` state.
- Derived state: the projection is read-only evidence. Terminal acknowledgements,
retention heads, filesystem artifacts, and pulse names cannot decide whether
a grant may launch.
- Forbidden writers: the tool adapter cannot discover sibling stores, inspect
generic CultCache cargo, mutate resident state, or infer authority from
filenames.
- Shared paths: Heartbeat issuance, pending-grant selection, and projection all
use `resident_self_grant_is_pending`; there is no observer-only definition of
launchability.
- Cut line: the native `epiphany_state.resident_grant_lifecycle` tool is exposed
only to Eyes, Modeling, and Soul and only when the authenticated resident
launch forwards its exact state-store binding.
- Verification: exact `34d75d05` unit coverage proves an older terminal grant
remains non-launchable beside a later retry. Exact packaged copied cognition
observed the same shape; after three strict duplicate-field result failures,
attempt four admitted typed evidence and resident Self terminalized the exact
original grant once under the brake. No new grant appeared.
The three failed attempts are not accepted reliability evidence. The next cut
belongs to role-result schema construction/model ingress: preserve strict
duplicate-field rejection while making the first model-produced result conform
to its one-owner schema.