@consiliency/agent-board-schema 1.23.0 → 1.25.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +75 -1
- package/fixtures/already-applied.json +12 -0
- package/migrations/20260905000100_agent_board_consent_gating_reconcile.sql +151 -0
- package/migrations/20260906000100_agent_board_enroll_resumption.sql +749 -0
- package/migrations/20260907000100_agent_board_signed_authority_provenance.sql +243 -0
- package/package.json +1 -1
- package/schema-manifest.json +30 -3
- package/src/check.js +23 -0
- package/src/index.js +20 -0
- package/src/plan.js +79 -5
- package/src/tooling.test.mjs +237 -4
- package/tests/agent_board_atomic_fleet_onboarding.test.sql +124 -112
- package/tests/agent_board_claimkey.test.sql +25 -3
- package/tests/agent_board_consent_emission.test.sql +17 -13
- package/tests/agent_board_consent_gating.test.sql +55 -11
- package/tests/agent_board_delegation_policies.test.sql +3 -1
- package/tests/agent_board_enroll_resumption.test.sql +412 -0
- package/tests/agent_board_federated_lock_order_class.test.sql +1 -3
- package/tests/agent_board_fedsession_provider_absence_gate.test.sql +9 -6
- package/tests/agent_board_fleet_machine_reactivation.test.sql +56 -5
- package/tests/agent_board_mint_harden.test.sql +39 -4
- package/tests/agent_board_outbox_wake_ordering.test.sql +27 -12
- package/tests/agent_board_rejectsem.test.sql +1 -1
- package/tests/agent_board_rolereset.test.sql +23 -2
- package/tests/agent_board_rowid.test.sql +24 -13
- package/tests/agent_board_scoped_send_regrant.test.sql +8 -4
- package/tests/agent_board_signed_authority_provenance.test.sql +454 -0
- package/tests/agent_board_source_strip_helper.test.sql +285 -0
- package/tests/agent_board_wake_recovery.test.sql +58 -40
- package/tests/catalog/p_metadata_parameters.json +689 -0
- package/tests/fixtures/test-prereqs.sql +313 -0
- package/tests/isolation/expected/atomic_fleet_lifecycle_races.out +3 -3
package/README.md
CHANGED
|
@@ -26,7 +26,7 @@ agent-board-schema materialize --target <adopter>/supabase/migrations
|
|
|
26
26
|
|
|
27
27
|
# Read-only posture against a supplied migration-history snapshot.
|
|
28
28
|
agent-board-schema plan --history history.json
|
|
29
|
-
agent-board-schema check --history history.json # exit 0 exact / 10 behind / 11 ahead / 12 divergent
|
|
29
|
+
agent-board-schema check --history history.json # exit 0 exact / 10 behind / 11 ahead / 12 divergent / 14 collision
|
|
30
30
|
|
|
31
31
|
# Self-contained smoke: byte-exact + idempotent materialize.
|
|
32
32
|
agent-board-schema materialize --check-smoke
|
|
@@ -35,6 +35,80 @@ agent-board-schema manifest # print the compatibility manifest
|
|
|
35
35
|
agent-board-schema version
|
|
36
36
|
```
|
|
37
37
|
|
|
38
|
+
## Version collisions — build `history.json` from real rows, not from versions
|
|
39
|
+
|
|
40
|
+
**Supabase keys `supabase_migrations.schema_migrations` on the 14-digit version
|
|
41
|
+
prefix alone, not the name.** If you have independently authored a migration at a
|
|
42
|
+
version this bundle also uses, your database has that version recorded under *your*
|
|
43
|
+
name and the bundle's migration is **skipped silently** — no error, and
|
|
44
|
+
`schema_version` still reports the bundle's number, so the deploy looks successful.
|
|
45
|
+
|
|
46
|
+
This is not hypothetical. It happened at `20260809000100`, where an adopter's
|
|
47
|
+
`reconcile_panel_schema_drift` occupied the version used by
|
|
48
|
+
`agent_board_consent_gating` (message-board#217). 50 of this bundle's migrations share
|
|
49
|
+
the `000100` suffix, so if you also mint fixed-suffix timestamps the odds are far
|
|
50
|
+
above random.
|
|
51
|
+
|
|
52
|
+
**Build the history from version *and* name**, so `check` can see it. Let Postgres
|
|
53
|
+
do the encoding — a shell pipeline breaks on a name containing `:` or a quote, and
|
|
54
|
+
an empty result set produces a file that will not parse:
|
|
55
|
+
|
|
56
|
+
```bash
|
|
57
|
+
psql "$DATABASE_URL" -At -c \
|
|
58
|
+
"select coalesce(json_agg(json_build_object('id', version || '_' || coalesce(name, ''))), '[]'::json)
|
|
59
|
+
from supabase_migrations.schema_migrations" \
|
|
60
|
+
| sed 's/^/{"applied":/; s/$/}/' > history.json
|
|
61
|
+
|
|
62
|
+
agent-board-schema check --history history.json
|
|
63
|
+
```
|
|
64
|
+
|
|
65
|
+
A history built from **versions alone** cannot detect this — matching on the version
|
|
66
|
+
prefix is precisely the check that reports success on an affected database.
|
|
67
|
+
|
|
68
|
+
`check` exits **14** and names both sides:
|
|
69
|
+
|
|
70
|
+
```
|
|
71
|
+
collision @20260809000100:
|
|
72
|
+
expected 20260809000100_agent_board_consent_gating
|
|
73
|
+
recorded as 20260809000100_reconcile_panel_schema_drift
|
|
74
|
+
```
|
|
75
|
+
|
|
76
|
+
Exit 14 is deliberately not `behind` (10). "Behind" invites the fix of applying
|
|
77
|
+
again, which is the no-op that leaves the database wrong.
|
|
78
|
+
|
|
79
|
+
**Single-version discriminator**, if you only want to check the known case:
|
|
80
|
+
|
|
81
|
+
```sql
|
|
82
|
+
select name from supabase_migrations.schema_migrations where version = '20260809000100';
|
|
83
|
+
-- agent_board_consent_gating -> healthy
|
|
84
|
+
-- anything else -> this database skipped the bundle migration
|
|
85
|
+
```
|
|
86
|
+
|
|
87
|
+
**Remediation.** From bundle version 1.24.0 you do not need one:
|
|
88
|
+
`20260905000100_agent_board_consent_gating_reconcile` re-applies the only function
|
|
89
|
+
left stale by that skip (`join_board` — the other two are superseded by later
|
|
90
|
+
migrations that carry the gating forward). It is byte-identical on a healthy
|
|
91
|
+
database and self-healing on an affected one, with no `supabase migration repair`.
|
|
92
|
+
|
|
93
|
+
**A healed database stops reporting `collision`, even though the row is permanent.**
|
|
94
|
+
`schema_migrations` keys on `version`, its primary key, so the adopter's row at a
|
|
95
|
+
collided version can never be removed. `check` therefore reports on whether the
|
|
96
|
+
*remediating migration is applied*, not on whether the row exists — otherwise a
|
|
97
|
+
database that had done everything right would gate red forever, and an adopter
|
|
98
|
+
trained to ignore exit 14 would stop seeing genuine `behind` and `divergent`
|
|
99
|
+
states too. A remediated collision is reported under `collisionsRemediated` and both halves of
|
|
100
|
+
the pair are removed from the counts, so a healed database reports **`exact` / exit
|
|
101
|
+
0**. Removing them is the whole point: the bundle's migration can never be applied
|
|
102
|
+
there (its version is taken) and your row can never be removed, so left in they would
|
|
103
|
+
report `behind` forever — a different permanent red than the one being fixed.
|
|
104
|
+
|
|
105
|
+
A row whose `name` is NULL is treated as **indeterminate**, not as a collision: it
|
|
106
|
+
carries no name to tell apart from the bundle's own migration at that version. The
|
|
107
|
+
same applies to a history built from bare version strings.
|
|
108
|
+
|
|
109
|
+
For a collision at any *other* version there is no declared healer, so `check`
|
|
110
|
+
keeps reporting it and the fix is a `migration repair` on your side.
|
|
111
|
+
|
|
38
112
|
After applying the materialized migrations, the adopter records the install so the
|
|
39
113
|
authenticated compatibility RPC reports live:
|
|
40
114
|
|
|
@@ -488,6 +488,18 @@
|
|
|
488
488
|
{
|
|
489
489
|
"id": "20260904000100_agent_board_wake_recovery_surface",
|
|
490
490
|
"sha256": "6349f705751104c5d60ec6602ea534e7c8b7de5c4c6ad2d84dbda680a5267848"
|
|
491
|
+
},
|
|
492
|
+
{
|
|
493
|
+
"id": "20260905000100_agent_board_consent_gating_reconcile",
|
|
494
|
+
"sha256": "d5328a0006ecd1ec940c4fde5bbb43acf29b0997eefbea557009f792868e298b"
|
|
495
|
+
},
|
|
496
|
+
{
|
|
497
|
+
"id": "20260906000100_agent_board_enroll_resumption",
|
|
498
|
+
"sha256": "2aafad9df5da006039268d9559311f85fba634b6ec44d06a2f55fe3deedb09d8"
|
|
499
|
+
},
|
|
500
|
+
{
|
|
501
|
+
"id": "20260907000100_agent_board_signed_authority_provenance",
|
|
502
|
+
"sha256": "21111e289e0bb70ac3236f5422476c0c51d0bca5a815c11dd4e48b6932cf7c6b"
|
|
491
503
|
}
|
|
492
504
|
]
|
|
493
505
|
}
|
|
@@ -0,0 +1,151 @@
|
|
|
1
|
+
-- Reconcile join_board onto its consent-gated definition (message-board#217).
|
|
2
|
+
--
|
|
3
|
+
-- WHY THIS EXISTS. Supabase keys supabase_migrations.schema_migrations on the
|
|
4
|
+
-- 14-digit VERSION PREFIX alone, not the name. An adopter that independently
|
|
5
|
+
-- authored a migration at 20260809000100 therefore has that version recorded
|
|
6
|
+
-- under its own name, and the bundle's 20260809000100_agent_board_consent_gating
|
|
7
|
+
-- is SKIPPED SILENTLY -- no error, because that file is three
|
|
8
|
+
-- `create or replace function` statements. The database reaches schema_version 8
|
|
9
|
+
-- while retaining pre-consent authorization semantics, and the deploy reports
|
|
10
|
+
-- success. Found by Consiliency/consiliency-portal while adopting 1.22.0.
|
|
11
|
+
--
|
|
12
|
+
-- WHY join_board AND NOTHING ELSE. The skipped file redefines three functions,
|
|
13
|
+
-- but two of them are superseded by LATER migrations that carry the consent
|
|
14
|
+
-- gating forward: resolve_board_request by 20260819000100 and
|
|
15
|
+
-- update_board_membership by 20260813000100. join_board has no later definition
|
|
16
|
+
-- -- its newest is the skipped one -- so on an affected database it is still the
|
|
17
|
+
-- pre-consent 20260803000100 body. That single function is the entire live gap.
|
|
18
|
+
--
|
|
19
|
+
-- WHY NOT RENUMBER THE ORIGINAL. Renumbering looks safe because the migration is
|
|
20
|
+
-- idempotent, but idempotent is not order-independent. On a HEALTHY database the
|
|
21
|
+
-- renumbered file would be the only unapplied migration, so it would run LAST --
|
|
22
|
+
-- after 20260813000100 and 20260819000100 -- and re-apply the older
|
|
23
|
+
-- resolve_board_request and update_board_membership bodies over the newer ones.
|
|
24
|
+
-- That is a silent authorization ROLLBACK on every correctly-adopted database,
|
|
25
|
+
-- including production. A forward-only migration cannot do that, because it runs
|
|
26
|
+
-- last by construction and carries the newest definition of the only function it
|
|
27
|
+
-- touches.
|
|
28
|
+
--
|
|
29
|
+
-- EFFECT. Healthy database: byte-identical body, a true no-op. Affected database:
|
|
30
|
+
-- receives the consent gating it never got, with no `supabase migration repair`
|
|
31
|
+
-- and no operator intervention. `create or replace` preserves the existing ACL
|
|
32
|
+
-- and this migration adds no new grants, so the privilege surface is unchanged
|
|
33
|
+
-- either way.
|
|
34
|
+
|
|
35
|
+
create or replace function agent_board.join_board(p_board_id uuid, p_agent_id uuid, p_membership_state text DEFAULT 'joined'::text, p_created_by_agent_id uuid DEFAULT NULL::uuid, p_metadata jsonb DEFAULT '{}'::jsonb)
|
|
36
|
+
RETURNS TABLE(membership_id uuid, board_id uuid, agent_id uuid, membership_state text, created_at timestamp with time zone, updated_at timestamp with time zone)
|
|
37
|
+
LANGUAGE plpgsql
|
|
38
|
+
SECURITY DEFINER
|
|
39
|
+
SET search_path TO 'agent_board', 'public'
|
|
40
|
+
as $$
|
|
41
|
+
declare
|
|
42
|
+
v_board agent_board.boards%rowtype;
|
|
43
|
+
v_existing_state text;
|
|
44
|
+
v_membership agent_board.board_memberships%rowtype;
|
|
45
|
+
v_actor uuid := agent_board.current_agent_id();
|
|
46
|
+
begin
|
|
47
|
+
-- GATENULL (#58): only service_role may proceed with an unresolved actor.
|
|
48
|
+
if v_actor is null and not agent_board._is_service_role() then
|
|
49
|
+
raise exception 'current agent principal not resolved' using errcode = '42501';
|
|
50
|
+
end if;
|
|
51
|
+
|
|
52
|
+
if p_membership_state not in ('joined', 'subscribed') then
|
|
53
|
+
raise exception 'join_board only accepts joined or subscribed' using errcode = '22023';
|
|
54
|
+
end if;
|
|
55
|
+
if not exists (select 1 from agent_board.agents where id = p_agent_id) then
|
|
56
|
+
raise exception 'agent not found' using errcode = 'P0002';
|
|
57
|
+
end if;
|
|
58
|
+
if p_created_by_agent_id is not null and not exists (select 1 from agent_board.agents where id = p_created_by_agent_id) then
|
|
59
|
+
raise exception 'creator agent not found' using errcode = 'P0002';
|
|
60
|
+
end if;
|
|
61
|
+
|
|
62
|
+
select * into v_board from agent_board.boards where id = p_board_id and status = 'active';
|
|
63
|
+
if v_board.id is null then
|
|
64
|
+
raise exception 'board not found' using errcode = 'P0002';
|
|
65
|
+
end if;
|
|
66
|
+
|
|
67
|
+
-- MBROWID (3a): prefer ACTIVE standing for joinability / idempotent-join. A bare
|
|
68
|
+
-- total-order can surface a high-id REMOVED row while a low-id JOINED row exists,
|
|
69
|
+
-- locking an active member out of their own private / invite_only board (P0002).
|
|
70
|
+
select bm.membership_state
|
|
71
|
+
into v_existing_state
|
|
72
|
+
from agent_board.board_memberships bm
|
|
73
|
+
where bm.board_id = p_board_id
|
|
74
|
+
and bm.agent_id = p_agent_id
|
|
75
|
+
order by
|
|
76
|
+
case when bm.membership_state in ('joined', 'subscribed') then 0 else 1 end,
|
|
77
|
+
bm.created_at desc,
|
|
78
|
+
bm.id desc
|
|
79
|
+
limit 1;
|
|
80
|
+
|
|
81
|
+
-- #82: the CREATOR EXEMPTION IS REMOVED. Previously a board's creator could join ANY agent
|
|
82
|
+
-- to their board -- `v_actor <> v_board.created_by_agent_id` let them past this guard, and
|
|
83
|
+
-- the joinability check below only runs for self-joins, so nothing else stopped them. That
|
|
84
|
+
-- is the force-join hole. Creating a board does not entitle you to place other people in it.
|
|
85
|
+
--
|
|
86
|
+
-- Only the subject may join the subject. service_role (v_actor null) keeps the operator
|
|
87
|
+
-- carve-out, which is deliberate and maintainer-confirmed.
|
|
88
|
+
if v_actor is not null and v_actor <> p_agent_id then
|
|
89
|
+
raise exception 'board membership actor mismatch' using errcode = '42501';
|
|
90
|
+
end if;
|
|
91
|
+
|
|
92
|
+
-- NULLFIX (CR): v_existing_state is NULL when the caller has NO membership row, and the original
|
|
93
|
+
-- guard then evaluated to NULL — `not (false or false or NULL)` is NULL, not TRUE — so `if NULL
|
|
94
|
+
-- then` never fired and ANY resolvable actor self-joined ANY board, private ones included. That
|
|
95
|
+
-- (not the null-actor bypass) was the true root cause of the live "joined a private invite_only
|
|
96
|
+
-- board with no invite" finding, and it is reachable by an ACTIVE agent, which the GATENULL guard
|
|
97
|
+
-- above cannot touch. coalesce() makes the absent-row case FALSE so the raise fires.
|
|
98
|
+
--
|
|
99
|
+
-- MBROWID (3b): a 'removed' agent is excluded from self-joining even an OPEN board via the
|
|
100
|
+
-- R-LINK-7 SET predicate (exists removed AND no active), not a single latest-row state.
|
|
101
|
+
-- Total-ordering the latest-row read alone is insufficient: a removed+left pair with equal
|
|
102
|
+
-- created_at can surface 'left' as the total-order winner while standing is still removed
|
|
103
|
+
-- (mutator + grant_board_role already use set-based rules). The three defenses must agree.
|
|
104
|
+
if v_actor is not null
|
|
105
|
+
and v_actor = p_agent_id
|
|
106
|
+
and v_board.created_by_agent_id <> v_actor
|
|
107
|
+
and not (
|
|
108
|
+
not (
|
|
109
|
+
exists (
|
|
110
|
+
select 1 from agent_board.board_memberships bm
|
|
111
|
+
where bm.board_id = p_board_id
|
|
112
|
+
and bm.agent_id = p_agent_id
|
|
113
|
+
and bm.membership_state = 'removed'
|
|
114
|
+
) and not exists (
|
|
115
|
+
select 1 from agent_board.board_memberships bm
|
|
116
|
+
where bm.board_id = p_board_id
|
|
117
|
+
and bm.agent_id = p_agent_id
|
|
118
|
+
and bm.membership_state in ('joined', 'subscribed')
|
|
119
|
+
)
|
|
120
|
+
)
|
|
121
|
+
and (
|
|
122
|
+
v_board.visibility = 'open'
|
|
123
|
+
-- #108: HOLDING AN INVITATION ADMITS THE SUBJECT ON ANY VISIBILITY, not only
|
|
124
|
+
-- invite_only. Previously `private` and `directory_hidden` were excluded outright, so
|
|
125
|
+
-- an invited subject had NO act available -- verified: a first-time invitee with no
|
|
126
|
+
-- prior refusal was blocked identically to a refused one. Gating makes consent
|
|
127
|
+
-- REQUIRED, so a board type where consent cannot be GIVEN is correct-but-unusable.
|
|
128
|
+
--
|
|
129
|
+
-- This is the narrowest widening that makes consent exercisable. It admits nobody new:
|
|
130
|
+
-- this whole guard is reached only when v_actor = p_agent_id (the actor-mismatch raise
|
|
131
|
+
-- above returns otherwise), so it is self-accept only. The board still decides who is
|
|
132
|
+
-- invited; the subject decides whether to accept.
|
|
133
|
+
or coalesce(v_existing_state, '') = 'invited'
|
|
134
|
+
or coalesce(v_existing_state, '') in ('joined', 'subscribed')
|
|
135
|
+
)
|
|
136
|
+
) then
|
|
137
|
+
raise exception 'board is not joinable by this agent' using errcode = 'P0002';
|
|
138
|
+
end if;
|
|
139
|
+
|
|
140
|
+
v_membership := agent_board._upsert_board_membership_state(
|
|
141
|
+
p_board_id,
|
|
142
|
+
p_agent_id,
|
|
143
|
+
p_membership_state,
|
|
144
|
+
coalesce(p_created_by_agent_id, p_agent_id),
|
|
145
|
+
p_metadata
|
|
146
|
+
);
|
|
147
|
+
|
|
148
|
+
return query
|
|
149
|
+
select v_membership.id, v_membership.board_id, v_membership.agent_id, v_membership.membership_state, v_membership.created_at, v_membership.updated_at;
|
|
150
|
+
end;
|
|
151
|
+
$$;
|