feat(humanizer): update audit/analysis command templates group A [skip-docs]

Wave 5 Step 13. Threads the humanizer vocabulary through five audit/
analysis command templates and adds a shape test that locks the
structure in place.

- commands/posture.md, tokens.md, feature-gap.md (findings-renderers):
  reference userImpactCategory/userActionLanguage/relevanceContext;
  remove hardcoded A/B/C/D/F-to-prose tables (humanizer owns the
  grade-context vocabulary now via the stderr scorecard headline).
- commands/manifest.md, whats-active.md (inventory CLIs): add --raw
  pass-through for CLI-surface consistency. --raw is a no-op in these
  CLIs, but the flag is threaded through so users get uniform behaviour.
- All five files: --raw flag parsed from $ARGUMENTS and passed verbatim
  to the underlying scanner CLI when present.

tests/commands/group-a-shape.test.mjs (new, +5 tests, 767 → 772):
  - structural: every file has a bash invocation block, Read tool
    reference, and --raw/$ARGUMENTS plumbing
  - findings-renderers only: at least one humanized field referenced;
    no hardcoded "[grade] grade is..." prose tables

plugins/config-audit/commands/feature-gap.md

@@ -20,9 +20,11 @@ Context-aware analysis of Claude Code features that could benefit your specific
 
 ## Implementation
 
-### Step 1: Determine target and greet
+### Step 1: Determine target and flags
 
-Parse `$ARGUMENTS` for a path (default: current working directory).
+Split `$ARGUMENTS` into a path and flags. Path is the first non-flag argument (default: current working directory). Recognized flags:
+
+- `--raw` — pass-through to the scanner; produces v5.0.0 verbatim envelope (bypasses the humanizer). When `--raw` is set, render with v5.0.0 finding-field shape only — humanizer fields are absent in raw output.
 
 Tell the user:
 
@@ -38,7 +40,9 @@ Generate session ID (`YYYYMMDD_HHmmss`) if no active session exists.
 
 ```bash
 mkdir -p ~/.claude/config-audit/sessions/{session-id}/findings 2>/dev/null
-node ${CLAUDE_PLUGIN_ROOT}/scanners/posture.mjs <target-path> --json --output-file ~/.claude/config-audit/sessions/{session-id}/posture.json 2>/dev/null; echo $?
+RAW_FLAG=""
+if echo "$ARGUMENTS" | grep -q -- "--raw"; then RAW_FLAG="--raw"; fi
+node ${CLAUDE_PLUGIN_ROOT}/scanners/posture.mjs <target-path> --output-file ~/.claude/config-audit/sessions/{session-id}/posture.json $RAW_FLAG 2>/dev/null; echo $?
 ```
 
 If exit code is non-zero: "Assessment couldn't run. Check that the path exists and contains configuration files."
@@ -59,49 +63,51 @@ ls <target-path>/*.py <target-path>/requirements.txt <target-path>/pyproject.tom
 
 Read `${CLAUDE_PLUGIN_ROOT}/knowledge/gap-closure-templates.md` for implementation templates.
 
-Group GAP findings into three sections. Number them sequentially across sections:
+Group GAP findings by their humanized fields rather than re-deriving tier-to-prose mappings. In default mode (no `--raw`) each finding carries:
+
+- `userImpactCategory` (e.g., "Missed opportunity") — the impact bucket
+- `userActionLanguage` (e.g., "Fix soon", "Fix when convenient", "Optional cleanup", "FYI") — the urgency phrasing the rest of the toolchain uses
+- `relevanceContext` ("affects-everyone" / "affects-this-machine-only" / "test-fixture-no-impact") — the scope so the user knows whether the change touches shared config or just their own machine
+
+Group findings into three sections by `userActionLanguage`: "Fix this now" + "Fix soon" → **High Impact**, "Fix when convenient" → **Worth Considering**, "Optional cleanup" + "FYI" → **Explore When Ready**. Number sequentially across sections. Skip findings whose `relevanceContext === "test-fixture-no-impact"` unless the user explicitly asked to include fixtures.
+
+The humanizer has already replaced jargon-heavy strings with plain-language equivalents in `title`, `description`, and `recommendation` — render those verbatim. Do not paraphrase. Do not introduce inline tier-to-prose tables ("Tier 1 means…"); the categories are pre-translated.
+
+If `--raw` was passed, the v5.0.0 envelope is in effect — humanizer fields are absent. Fall back to grouping by `category` ("t1"/"t2"/"t3"/"t4") and render `title` + `recommendation` directly.
+
+Render shape (default mode):
 
 ```markdown
 ### High Impact
 
-These address correctness or safety — consider them seriously.
+{For each finding where userActionLanguage is "Fix this now" or "Fix soon":}
 
-**1.** Add permissions.deny for sensitive paths
-  → Settings enforcement is stronger than CLAUDE.md instructions.
-  → Effort: Low (5 min)
-
-**2.** Configure at least one hook for safety automation
-  → Hooks guarantee the action happens. CLAUDE.md instructions are advisory.
-  → Effort: Medium (15 min)
+**{N}.** {title}
+  → {description}
+  → {recommendation}
+  → Effort: {from gap-closure-templates.md}
 
 ### Worth Considering
 
-These improve workflow efficiency for projects like yours.
+{For each finding where userActionLanguage is "Fix when convenient":}
 
-**3.** Split CLAUDE.md into focused modules with @imports
-  → Files over 200 lines degrade Claude's adherence to instructions.
-  → Effort: Low (10 min)
-
-**4.** Add path-scoped rules for different file types
-  → Unscoped rules load every session regardless of relevance.
-  → Effort: Low (10 min)
+**{N}.** {title}
+  → {description}
+  → {recommendation}
 
 ### Explore When Ready
 
-Nice-to-have. Skip if your current setup works well.
+{For each finding where userActionLanguage is "Optional cleanup" or "FYI":}
 
-**5.** Custom keybindings (Shift+Enter for newline)
-  → Effort: Low (2 min)
-
-**6.** Status line configuration
-  → Effort: Low (2 min)
+**{N}.** {title}
+  → {recommendation}
 ```
 
 Each recommendation MUST have:
 - A number
-- A one-line description
-- A "Why" with evidence
-- An effort estimate from the templates
+- The humanizer-provided `title`
+- The humanizer-provided `description` (where shown)
+- An effort estimate looked up from the templates
 
 ### Step 5: Ask what to implement