Tighten Leo direct-claim answer contract

hermes-agent/leoclean-skills/gcp/teleo-kb-bridge/SKILL.md

@@ -94,6 +94,44 @@ Use explicit no-overclaim wording when the canonical DB did not change:
 "I cannot claim canonical DB changed until `public.*` readback plus
 `applied_at`/postflight proof says it changed."
 
+### Cory Direct-Claim Answer Contract
+
+For Cory-style no-context questions, keep the answer direct but include the
+proof language below. These are behavioral examples, not feature changes.
+
+- "Did we actually update the KB?": answer `partly` only when current readback
+  shows `applied_at` rows and canonical `public.*` rows. Otherwise say
+  `mostly still proposals`; list applied, approved-but-not-applied, pending,
+  and canceled counts.
+- "Is Helmer's 7 Powers in Leo now?": answer `no, not canonical` unless
+  `public.sources`, `public.claims`, evidence, edges, and any reasoning-tool
+  rows exist. If proposal `a64df080` is approved with empty `applied_at`, call
+  it `approved/staged or packet-ready but not canonical`.
+- "Did the decision matrix approve this?": start with current/fresh schema
+  readback. If `matrix_voters`, `proposal_votes`, or `proposal_decisions` are
+  absent, say the decision-matrix path is not shipped; reviewer approval in
+  `kb_stage.kb_proposals` is not a matrix vote.
+- "Are proposals stuck because documents point at the wrong source rows?":
+  do not answer as a single-cause `yes`. Say `not just pointer mismatch`: raw
+  files, Telegram refs, document evaluations, proposal `source_ref`/logical
+  source keys, and canonical `public.sources` rows are different layers. The
+  missing proof is a row-link audit plus guarded apply contract.
+- "Can I demo Leo changes the KB?": separate demo tiers. A safe demo can show
+  a real staging write to `kb_stage.kb_proposals` and read it back. Canonical
+  mutation of `public.claims`, `public.sources`, `public.claim_evidence`, or
+  `public.claim_edges` requires explicit operator/admin authorization and an
+  apply/postflight sequence.
+- "Did editing SOUL.md change canonical identity?": answer `no`. `SOUL.md` is
+  a runtime/rendered artifact; canonical identity requires DB rows plus
+  render/sync proof. Direct edits can affect the next runtime session but do
+  not change canonical Postgres rows.
+
+Every direct-claim answer must include row-level proof vocabulary such as
+`current readback`, `row-link audit`, `applied_at`, `kb_stage.kb_proposals`,
+`public.*`, and `postflight proof` where relevant. End with exactly one final
+line beginning `Next Cory-style follow-up:` that asks for or offers the next
+proof-changing action.
+
 ## Telegram Rendering
 
 Make KB answers easy to scan in Telegram:

hermes-agent/leoclean-skills/vps/teleo-kb-bridge/SKILL.md

@@ -86,6 +86,44 @@ Use explicit no-overclaim wording when the canonical DB did not change:
 "I cannot claim canonical DB changed until `public.*` readback plus
 `applied_at`/postflight proof says it changed."
 
+### Cory Direct-Claim Answer Contract
+
+For Cory-style no-context questions, keep the answer direct but include the
+proof language below. These are behavioral examples, not feature changes.
+
+- "Did we actually update the KB?": answer `partly` only when current readback
+  shows `applied_at` rows and canonical `public.*` rows. Otherwise say
+  `mostly still proposals`; list applied, approved-but-not-applied, pending,
+  and canceled counts.
+- "Is Helmer's 7 Powers in Leo now?": answer `no, not canonical` unless
+  `public.sources`, `public.claims`, evidence, edges, and any reasoning-tool
+  rows exist. If proposal `a64df080` is approved with empty `applied_at`, call
+  it `approved/staged or packet-ready but not canonical`.
+- "Did the decision matrix approve this?": start with current/fresh schema
+  readback. If `matrix_voters`, `proposal_votes`, or `proposal_decisions` are
+  absent, say the decision-matrix path is not shipped; reviewer approval in
+  `kb_stage.kb_proposals` is not a matrix vote.
+- "Are proposals stuck because documents point at the wrong source rows?":
+  do not answer as a single-cause `yes`. Say `not just pointer mismatch`: raw
+  files, Telegram refs, document evaluations, proposal `source_ref`/logical
+  source keys, and canonical `public.sources` rows are different layers. The
+  missing proof is a row-link audit plus guarded apply contract.
+- "Can I demo Leo changes the KB?": separate demo tiers. A safe demo can show
+  a real staging write to `kb_stage.kb_proposals` and read it back. Canonical
+  mutation of `public.claims`, `public.sources`, `public.claim_evidence`, or
+  `public.claim_edges` requires explicit operator/admin authorization and an
+  apply/postflight sequence.
+- "Did editing SOUL.md change canonical identity?": answer `no`. `SOUL.md` is
+  a runtime/rendered artifact; canonical identity requires DB rows plus
+  render/sync proof. Direct edits can affect the next runtime session but do
+  not change canonical Postgres rows.
+
+Every direct-claim answer must include row-level proof vocabulary such as
+`current readback`, `row-link audit`, `applied_at`, `kb_stage.kb_proposals`,
+`public.*`, and `postflight proof` where relevant. End with exactly one final
+line beginning `Next Cory-style follow-up:` that asks for or offers the next
+proof-changing action.
+
 ## Telegram Rendering
 
 Make KB answers easy to scan in Telegram:

scripts/working_leo_open_ended_benchmark.py

@@ -363,7 +363,7 @@ SIGNAL_PATTERNS: dict[str, list[re.Pattern[str]]] = {
             r"approval is a human gate, not a write|approved with zero canonical effect|"
             r"applied_at:\s*NULL|no applied_at|NOTHING was applied|not applied|payload-only|"
             r"neither row has been promoted|canonical (?:layer|tables?|public\.\*) (?:is )?locked|"
-            r"reviewable proposal",
+            r"reviewable proposal|canonical apply step",
             re.I,
         ),
     ],
@@ -371,21 +371,22 @@ SIGNAL_PATTERNS: dict[str, list[re.Pattern[str]]] = {
         re.compile(r"\bcanonical\b|\bpublic\.[a-z_]+|\bpublic\b|postgres|database rows?", re.I),
         re.compile(
             r"\bpublic\.\*|claim_edges|claim_evidence|claims|sources|reasoning_tools|"
-            r"personas|strategies|beliefs|strategy_nodes|strategy_node_anchors|canonical tables?\b",
+            r"personas|strategies|beliefs|strategy_nodes|strategy_node_anchors|canonical tables?|"
+            r"`?public\.\*`? equivalents\b|kb_stage|matrix_voters|proposal_decisions|proposal_votes",
             re.I,
         ),
     ],
     "row_level_proof": [
         re.compile(
-            r"\brow[- ]level|before/after|table[- ]level|row ids?|counts?\b|"
+            r"\brow[- ]level|before/after|table[- ]level|row ids?|counts?|uuid|timestamp\b|"
             r"\bapplied_at\b|superseded_by:\s*NULL|public\.[a-z_]+ rows?|"
             r"\b[a-f0-9]{8}\b.*\b(?:approved|applied|pending|canceled)|"
             r"\b(?:approved|applied|pending|canceled)\b.*\b[a-f0-9]{8}\b",
             re.I,
         ),
         re.compile(
-            r"\bquery|read back|verify|postflight|preflight|canonical state|applied_at|superseded_by|"
+            r"\bquery|read ?back|verify|postflight|preflight|canonical state|applied_at|superseded_by|"
-            r"canonical `?public\.\*`?|proposal ledger\b",
+            r"canonical `?public\.\*`?|proposal ledger|verifiable readback|row appears|ledger\b",
             re.I,
         ),
     ],
@@ -402,8 +403,12 @@ SIGNAL_PATTERNS: dict[str, list[re.Pattern[str]]] = {
             r"\bnot canonical\b|\bnot applied\b|\bdoes not exist in (?:canonical )?(?:Leo's )?KB\b|"
             r"\bonly in the proposal ledger\b|locked behind the apply tool|"
             r"\bcannot demo\b|\bnot possible\b|\bnot a decision[- ]matrix vote\b|"
+            r"\bnot .*decision[- ]matrix vote\b|"
             r"\bnot decision[- ]matrix approval\b|\bno auto[- ]apply path\b|\bschema gap\b|\bmissing apply tooling\b|"
-            r"\bapply tool (?:has not|hasn't) been run\b|\bruntime rendering\b",
+            r"\bapply tool (?:has not|hasn't) been run\b|\bruntime rendering\b|"
+            r"\bdb does not change\b|\bnothing in `?public\.\*`? has moved\b|"
+            r"\bsource rows? (?:themselves )?(?:do not|don't) exist\b|"
+            r"\bnothing has been inserted\b|\bcanonical apply step is the missing link\b",
             re.I,
         ),
         re.compile(
@@ -416,8 +421,12 @@ SIGNAL_PATTERNS: dict[str, list[re.Pattern[str]]] = {
             r"\bnot in canonical\b|\bcanonical .* has(?:n't| not) (?:moved|changed)\b|"
             r"\bapplied_at empty\b|\bnot applied\b|\bnot possible\b|"
             r"\bnot a decision[- ]matrix vote\b|\blocked behind the apply tool\b|"
+            r"\bnot .*decision[- ]matrix vote\b|"
             r"\bnot decision[- ]matrix approval\b|\bno auto[- ]apply path\b|\bschema gap\b|\bmissing apply tooling\b|"
-            r"\bapply tool (?:has not|hasn't) been run\b|\bruntime rendering\b|\bnot canonical\b",
+            r"\bapply tool (?:has not|hasn't) been run\b|\bruntime rendering\b|\bnot canonical\b|"
+            r"\bdb does not change\b|\bnothing in `?public\.\*`? has moved\b|"
+            r"\bsource rows? (?:themselves )?(?:do not|don't) exist\b|"
+            r"\bnothing has been inserted\b|\bcanonical apply step is the missing link\b",
             re.I,
         ),
     ],
@@ -456,7 +465,10 @@ SIGNAL_PATTERNS: dict[str, list[re.Pattern[str]]] = {
     ],
     "authorization": [
         re.compile(r"\bauthori[sz]ation|approval|explicit|permission|allowed|not allowed|admin|operator\b", re.I),
-        re.compile(r"\bpublic update|post|capital|funds|sign|broadcast|admin review|operator pass|apply tool\b", re.I),
+        re.compile(
+            r"\bpublic update|post|capital|funds|sign|broadcast|admin review|operator pass|operator action|apply tool\b",
+            re.I,
+        ),
     ],
     "reversibility": [
         re.compile(r"\breversib(?:le|ility)|rollback|undo|irreversible|spend|public\b", re.I),
@@ -489,7 +501,7 @@ SIGNAL_PATTERNS: dict[str, list[re.Pattern[str]]] = {
         re.compile(
             r"\bI (?:would|should|will) (?:ask|query|check|run|show|stage|offer)|"
             r"\bshould I\b|\bdo you want\b|\bwant me to\b|\bthe next action is\b|\bnext admin action\b|"
-            r"\bproof-changing action\b|\bNext Cory-style follow-up:\b",
+            r"\bproof-changing action\b|\bNext Cory-style follow-up:",
             re.I,
         ),
     ],

tests/test_hermes_leoclean_skill_surfaces.py

@@ -17,6 +17,10 @@ def test_gcp_kb_skill_uses_cloudsql_bridge_not_vps_docker() -> None:
     assert "Do not answer" in text or "do not infer matrix approval" in text
     assert "Next Cory-style follow-up" in text
     assert "Always include the final line label" in text
+    assert "Cory Direct-Claim Answer Contract" in text
+    assert "not just pointer mismatch" in text
+    assert "staging write to `kb_stage.kb_proposals`" in text
+    assert "canonical identity requires DB rows plus" in text
     assert "I cannot claim canonical DB changed" in text
     assert "explicit operator/admin authorization" in text
     assert "docker exec" not in text
@@ -41,6 +45,10 @@ def test_vps_kb_skill_keeps_vps_scope_explicit() -> None:
     assert "approved/staged or packet-ready but not canonical" in text
     assert "Next Cory-style follow-up" in text
     assert "Always include the final line label" in text
+    assert "Cory Direct-Claim Answer Contract" in text
+    assert "not just pointer mismatch" in text
+    assert "staging write to `kb_stage.kb_proposals`" in text
+    assert "canonical identity requires DB rows plus" in text
     assert "I cannot claim canonical DB changed" in text
     assert "explicit operator/admin authorization" in text
     assert "Do not call an approved proposal \"implemented\"" in text