krwhynot · krwhynot · Aug 3, 2026 · Aug 3, 2026 · Aug 3, 2026 · Aug 3, 2026
diff --git a/docs/plans/2026-07-20-plugin-hardening-v5/changes/CR-6.md b/docs/plans/2026-07-20-plugin-hardening-v5/changes/CR-6.md
@@ -3,8 +3,18 @@
 **Raised:** 2026-07-30, after Codex round 4 (`gpt-5.6-sol` @ `xhigh`, session `019fb495`)
 **Raised by:** Claude (optimizer), on the reviewer's adjudication of a conflict I had escalated
 rather than resolved
-**State: PROPOSED** — needs owner ratification. **Nothing is applied.** F30 stays open and out of
-`findings_closed` until this is accepted and implemented.
+**State: RATIFIED** — owner ratification 2026-08-03; implemented the same day. The F30 guard
+in `tests/layer1-repo.sh` now enforces both halves: semantic absence (the sweep covers every
+tracked `.md` with NO allowlist, and SPLIT-3's resolver independently fails any shipping
+reference that cannot resolve) and the occurrence-addressed ledger
+(`tests/fixtures/f30-provenance-ledger.tsv`, 39 entries across 10 files at ratification).
+One implementation decision within the proposal's frame: occurrence address = file +
+sha256 of the exact line, NOT a file hash — a file-level hash would freeze every record
+containing one mention, punishing the truthful-rewording practice CR-4 established.
+All six acceptance probes executed and caught/passed as specified, including the one the
+allowlist could never catch (a stale instruction added inside this plan's own directory)
+and both controls. F30 accordingly moves into `findings_closed` and PHV5-053 into
+`tickets_complete`.
 **Replaces:** Wave 5 acceptance row 3, verbatim:
 
 > - [ ] F30 stale-reference sweep — no `/deepgrade:doc` or `commands/doc.md` string survives anywhere [G]

diff --git a/docs/plans/2026-07-20-plugin-hardening-v5/research/layer7-runtime-evidence.md b/docs/plans/2026-07-20-plugin-hardening-v5/research/layer7-runtime-evidence.md
@@ -82,18 +82,29 @@ its message claims.
 
 ### Prerequisite before running A–I
 
+> **Procedure refreshed 2026-08-03 (post-split, catalog at v7.0.0).** The plugin split (7.0.0)
+> moved the hooks under test: the five safety handlers (both PreToolUse, both PostToolUse, Stop)
+> now ship in `deepgrade-guard`; `deepgrade` retains SessionStart, SubagentStop and PreCompact.
+> Checks B–E and G below therefore exercise `deepgrade-guard`'s handlers; A and F exercise
+> `deepgrade`'s. The slots below remain EMPTY — nothing about the observations themselves is
+> carried over or inferred. `tests/layer7-runtime-proof.sh --manual` prints this same checklist
+> and was retargeted with the split.
+
 ```
 /plugin marketplace update deepgrade-marketplace
 /plugin update deepgrade
+/plugin update deepgrade-guard
 /reload-plugins
-/plugin details deepgrade        -> must report Hooks (6) incl. SubagentStop
+/plugin details deepgrade        -> must report Hooks (3) incl. SubagentStop
+/plugin details deepgrade-guard  -> must report Hooks (5)
 ```
 
-> **Changed as of `df8ac58`.** The catalog now pins a GitHub source object at `9f16277`, so these
-> commands install **from the pushed SHA, not from this working copy**. That is currently harmless —
-> hook code is byte-identical between `9f16277` and `df8ac58` — but it stops being harmless the moment
-> you edit anything under `hooks/` or `scripts/` and re-run this checklist. Push first, or the copy
-> under test is not the copy you changed.
+> The catalog pins a GitHub source object (v7.0.0 @ `55dbdeb`), so these commands install
+> **from the pinned SHA, not from this working copy**. That is harmless only while hook code is
+> byte-identical between the pin and HEAD — it stops being harmless the moment you edit anything
+> under `plugins/deepgrade/scripts/`, `plugins/deepgrade-guard/scripts/` or either `hooks/` and
+> re-run this checklist. Release (or test via `--plugin-dir` against the working copy) first, or
+> the copy under test is not the copy you changed.
 
 ### A. SessionStart (F26; settles part of U4)
 

diff --git a/docs/plans/2026-07-20-plugin-hardening-v5/status.json b/docs/plans/2026-07-20-plugin-hardening-v5/status.json
@@ -208,7 +208,8 @@
         "PHV5-041",
         "PHV5-042",
         "PHV5-043",
-        "PHV5-051"
+        "PHV5-051",
+        "PHV5-053"
       ],
       "findings_closed": [
         "F01",
@@ -229,7 +230,8 @@
         "F32",
         "F11",
         "F13",
-        "F14"
+        "F14",
+        "F30"
       ],
       "wave_1_complete": {
         "date": "2026-07-29",

diff --git a/interop.md b/interop.md
@@ -65,10 +65,12 @@ reproduced the documented shape instead of the actual one and hid the
 vocabulary drift above; provenance from a live artifact is what makes it a
 known-positive. One observed key was deliberately NOT ratified: a top-level
 `weight_set` appeared in the live artifact but in no template or schema and
-has no consumer; it stays out of the contract. Observed but unresolved: live
-elements also omit the `confidence` field some templates request — a
-semantic question (the scoring skill defines confidence levels) left open
-rather than silently ratified.
+has no consumer; it stays out of the contract. The per-check `confidence` field the templates
+originally requested was RESOLVED the same way (2026-08-03): recon showed it
+had no consumer anywhere — the report generator never reads it, and the
+methodology's "confidence levels" are a different, module-level concept the
+orchestrator derives from gate results, not from a checks[] field. The dead
+field is removed from the templates; module-level confidence is untouched.
 
 ## Deliberate non-edges
 

diff --git a/plugins/deepgrade-readiness/agents/budget-scanner.md b/plugins/deepgrade-readiness/agents/budget-scanner.md
@@ -429,7 +429,7 @@ Write results as JSON to docs/audit/readability/context-budget-scan.json followi
 the standard scanner output schema with all 8 checks.
 
 Each check must include:
-- id, name, status, points, max, confidence, evidence, details, remediation
+- id, name, status, points, max, evidence, details, remediation
 
 **Remediation guidance per check:**
 - 8.1 high tokens: "Move detailed instructions to .claude/rules/ with paths: or globs: frontmatter. Use progressive disclosure: CLAUDE.md points to docs, Claude reads on-demand."

diff --git a/plugins/deepgrade-readiness/agents/context-scanner.md b/plugins/deepgrade-readiness/agents/context-scanner.md
@@ -267,7 +267,7 @@ Write results as JSON to docs/audit/readability/context-scan.json following the
 standard scanner output schema with all 10 checks.
 
 Each check must include:
-- id, name, status, points, max, confidence, evidence, details, remediation
+- id, name, status, points, max, evidence, details, remediation
 
 **Constraints:**
 - Read-only. Do not modify any source files.

diff --git a/plugins/deepgrade-readiness/agents/database-scanner.md b/plugins/deepgrade-readiness/agents/database-scanner.md
@@ -590,7 +590,6 @@ standard scanner output schema.
       "status": "pass|partial|fail|skipped",
       "points": 0,
       "max": 3,
-      "confidence": "high|medium|low",
       "evidence": "...",
       "details": "...",
       "remediation": "..."
@@ -599,7 +598,7 @@ standard scanner output schema.
 }
 ```
 
-Each check must include: id, name, priority, status, points, max, confidence,
+Each check must include: id, name, priority, status, points, max,
 evidence, details, remediation.
 
 Status mapping:

diff --git a/plugins/deepgrade-readiness/agents/manifest-scanner.md b/plugins/deepgrade-readiness/agents/manifest-scanner.md
@@ -78,7 +78,6 @@ Do NOT change any "id" or "name" field. The check names are a contract.
       "status": "pass|fail",
       "points": <0 or 3>,
       "max": 3,
-      "confidence": "high|medium|low",
       "evidence": "<what was found>",
       "details": { "manifest_path": "<path>", "all_manifests_found": ["<paths>"] },
       "remediation": null
@@ -89,7 +88,6 @@ Do NOT change any "id" or "name" field. The check names are a contract.
       "status": "pass|warn|fail",
       "points": <0-2>,
       "max": 2,
-      "confidence": "high|medium|low",
       "evidence": "<what sections exist>",
       "details": { "has_name": <bool>, "has_deps": <bool>, "has_scripts": <bool> },
       "remediation": "<what's missing>"
@@ -100,7 +98,6 @@ Do NOT change any "id" or "name" field. The check names are a contract.
       "status": "pass|warn|fail",
       "points": <0-2>,
       "max": 2,
-      "confidence": "high|medium|low",
       "evidence": "<how determined>",
       "details": { "language": "<str>", "framework": "<str or null>", "method": "<manifest|deps|extension>" },
       "remediation": null
@@ -111,7 +108,6 @@ Do NOT change any "id" or "name" field. The check names are a contract.
       "status": "pass|warn|fail",
       "points": <0-2>,
       "max": 2,
-      "confidence": "high|medium|low",
       "evidence": "<summary of README content>",
       "details": { "readme_path": "<path or null>", "has_description": <bool>, "has_setup": <bool>, "has_run": <bool> },
       "remediation": "<what to add>"

diff --git a/tests/fixtures/f30-provenance-ledger.tsv b/tests/fixtures/f30-provenance-ledger.tsv
@@ -0,0 +1,47 @@
+# F30 provenance ledger (CR-6) — every intentional historical mention of the
+# deleted doc command, addressed by file + sha256 of the exact line content
+# (\r-stripped). The F30 guard fails on any occurrence NOT listed here, and on
+# any entry whose file no longer contains a matching line. Line-content hashes,
+# not file hashes: records that contain one mention stay editable elsewhere,
+# which is what the CR-4 truthful-rewording practice requires.
+# Registering a NEW entry is a provenance claim — it belongs in a change
+# record or a plan's own history, never in the same commit as a feature.
+docs/plans/2026-07-20-plugin-hardening-v5/acceptance-matrix.md	e6853e0a8d0feeca0cac0dd058a05c92fcf06fbd7e60b7f47f421603c229227d
+docs/plans/2026-07-20-plugin-hardening-v5/approach.md	77f3ff950477dc077571b5e95536b1fda50416f6fc7474ef862f85d88b4786a2
+docs/plans/2026-07-20-plugin-hardening-v5/approach.md	61e980dcea61b78c03aa968bfc09ab4aed853a09fc8bb06e1d3590daf5903369
+docs/plans/2026-07-20-plugin-hardening-v5/changes/CR-4.md	3d301f87992072978cfb5d3e9d68651987267fd128b82dd214b597ae6d830414
+docs/plans/2026-07-20-plugin-hardening-v5/changes/CR-4.md	7dee6b9416e47f84d3ba8070f83a6cf2d1a5ffc0d018399128166009b94fbda5
+docs/plans/2026-07-20-plugin-hardening-v5/changes/CR-4.md	7323964a02cae50a620f54fcdf9570dc873bc05958abf7bdc7df5eab6c5d406a
+docs/plans/2026-07-20-plugin-hardening-v5/changes/CR-4.md	982d1d4ab8eb247102b5b7ee463cc5824e6c3b9482de45e8c1002e496ebd05df
+docs/plans/2026-07-20-plugin-hardening-v5/changes/CR-4.md	c67934f2b5f2863c44b00a176afa60a1607c884140cb573237155e25c5318dba
+docs/plans/2026-07-20-plugin-hardening-v5/changes/CR-4.md	36a2a4338f77c9fc33d2e284037c18ee0ecb221b385219af901f9a34f10d362c
+docs/plans/2026-07-20-plugin-hardening-v5/changes/CR-4.md	a51f1525ab49dc4d3939b116a8ac849e7019f8a2524de34a4dab2d1820fe9452
+docs/plans/2026-07-20-plugin-hardening-v5/changes/CR-4.md	335603b911112447452b031a177fc7044f31fa7d5ee7a0c26ab6a6964a97f972
+docs/plans/2026-07-20-plugin-hardening-v5/changes/CR-4.md	506dcbe880722d979fa1c76ae920ee6adef19aa49f4cd65d7a138219df40f58d
+docs/plans/2026-07-20-plugin-hardening-v5/changes/CR-4.md	f6a77c6959030a42fabb03e048ef101088e49757207be7fb0902fa2b9ea835fd
+docs/plans/2026-07-20-plugin-hardening-v5/changes/CR-4.md	ff9dd134d04abe138a7ba5adbc8af9bb7fea18e2c846834275566854b8155c44
+docs/plans/2026-07-20-plugin-hardening-v5/changes/CR-4.md	77684a6230f46967fc2ee3eadc49f00c8be62228869b099f687363d98a9e81fc
+docs/plans/2026-07-20-plugin-hardening-v5/changes/CR-4.md	5b1eceac26a57f385fb6cc708d3f3cb3fe1a56ce9205b586badc8dae30118dbb
+docs/plans/2026-07-20-plugin-hardening-v5/changes/CR-6.md	7dee6b9416e47f84d3ba8070f83a6cf2d1a5ffc0d018399128166009b94fbda5
+docs/plans/2026-07-20-plugin-hardening-v5/changes/CR-6.md	65aac0f3799ea409530500b62874bcf3dada028be76af327a3d963ef8c2f5a4c
+docs/plans/2026-07-20-plugin-hardening-v5/codex-wave5-audit-prompt.md	c505fd57059eb3b2ec6608920a38a809948839c65e0c8cfd17179309e1181cd4
+docs/plans/2026-07-20-plugin-hardening-v5/codex-wave5-round4-prompt.md	ba7d59ca6b4fc046714190e26d3f3bc5fa0cff4cd2bad1b1f2910656feb341fe
+docs/plans/2026-07-20-plugin-hardening-v5/codex-wave5-round4-prompt.md	7c299a3ab04f0ec102e4d624568f4ab12310f852d23778b8d0e9e04c0acb3e33
+docs/plans/2026-07-20-plugin-hardening-v5/confidence.md	9f9ff55063c411826ebef8eedc8d4d3c829441edf2b36f6bd02b5c2f7e63b35a
+docs/plans/2026-07-20-plugin-hardening-v5/research/best-practices.md	2400e6c61184dfacfd205352d8d4d2f674f63fff22722a7eb697440a598f3676
+docs/plans/2026-07-20-plugin-hardening-v5/research/best-practices.md	4a99c57fcd0892a68e4762771ebb0af8e841da7a348fb19fb874fcae95235209
+docs/plans/2026-07-20-plugin-hardening-v5/research/best-practices.md	540f775e8ea452f49cd8bea0dce0991afc896421ca496f6f05671498ca342b34
+docs/plans/2026-07-20-plugin-hardening-v5/research/best-practices.md	200fe30a67acda51178ff5df09376407b942cb8aadfae5cdf85f34c40b0652db
+docs/plans/2026-07-20-plugin-hardening-v5/research/best-practices.md	a86a422185766e80a4869b43d323f6732dc13c12d70f19ee7c28200adc2a2669
+docs/plans/2026-07-20-plugin-hardening-v5/research/codebase-scan.md	3a8f182183ad6de5ed815f923e9ecb827ff8afc783767fadceffd4e4e3ce4b6d
+docs/plans/2026-07-20-plugin-hardening-v5/research/codebase-scan.md	a4178ef6ae8d411a9e07d8d7a36ceeeb255e76bedcbc8e0f9cea12baeb5342b4
+docs/plans/2026-07-20-plugin-hardening-v5/research/codebase-scan.md	7cd93c83ef7781dea6503d536f076414a69613fac41e7c12893a9b565ea974c9
+docs/plans/2026-07-20-plugin-hardening-v5/research/codebase-scan.md	0b23e4b82af736f653ce0fa9dcd044f6bc0cc34d3cc1ed0c552a30a7bc04dbb0
+docs/plans/2026-07-20-plugin-hardening-v5/research/codebase-scan.md	21231c01ac6b12797054f2965e8c0e2143e76945e1d26adbeaa5aa1a94afaff6
+docs/plans/2026-07-20-plugin-hardening-v5/research/codebase-scan.md	3c1fa4458ad8b0faa72f3c3b1ccca720e29cabc8a3f17c1bfc270c43a62b1b12
+docs/plans/2026-07-20-plugin-hardening-v5/research/codebase-scan.md	f004b555362aa375778b0de0e4e5dfbe06b9d64926554f9379b4c0329577f43d
+docs/plans/2026-07-20-plugin-hardening-v5/research/codebase-scan.md	f31645295d9a3dad0849718992fe514153701565970512808c3a12868a5bf908
+docs/plans/2026-07-20-plugin-hardening-v5/research/codebase-scan.md	7cc69bb6b4c464839ce50571f644adc47c301708c0478fcd263cf093b831c6e0
+docs/specs/plugin-hardening-v5.md	2e16326a4c37f4a673a76eaef4859b0508848af028073ceec974dc5505e16a2e
+docs/specs/plugin-hardening-v5.md	6d2df358afd5a639de92e59ad7efc6d2a2ee35593b78456f07ab22a8a2c6eb79
+docs/specs/plugin-hardening-v5.md	cc81f0c810473f344b305faf57faa928de64363476367c08c667ee488eb5628a