feat(plugin): after-write hook — PostToolUse on Bash diffs the working tree and runs the prior-art + ledger arms on what was just written; shared scribe_defs.sh; plugin 0.1.39 (#2901, milestone 299 step 3)

Edits made through sed/heredocs/scripts never reached the PreToolUse
Write|Edit hook, so a whole class of writes got no prior-art hint, no
ledger feed and no duplicate-family warning. scribe_after_write.sh asks git
what changed since it last looked (per-session path+blob snapshot; first
call = files touched in the last minute), extracts the definitions in the
added lines and calls /api/plugin/prior-art with the same three dedup
channels the pre hook keeps. Never blocks; silent on any failure. The
extractor, the prose/data skip list and the local by-name arm move to
scribe_defs.sh, sourced by both hooks. Version bump covers the step-2 hook
change too (run 4239 failed only on the bump check).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

plugin/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "scribe",
   "description": "Scribe system-of-record for Claude Code: MCP tools over your notes/tasks/projects/rules, a session-start push channel that surfaces your always-on rules + active-project context, process-skills (writing-plans, systematic-debugging, verification, brainstorming, reusing-code), and your saved Scribe Processes auto-surfaced as skills (/scribe:sync). Replaces superpowers + file-memory with one app-backed plugin.",
-  "version": "0.1.38",
+  "version": "0.1.39",
   "author": { "name": "Bryan Van Deusen" },
   "mcpServers": {
     "scribe": {

plugin/hooks/hooks.json

@@ -34,6 +34,17 @@
           }
         ]
       }
+    ],
+    "PostToolUse": [
+      {
+        "matcher": "Bash",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "bash \"${CLAUDE_PLUGIN_ROOT}/hooks/scribe_after_write.sh\""
+          }
+        ]
+      }
     ]
   }
 }

plugin/hooks/scribe_after_write.sh

@@ -0,0 +1,216 @@
+#!/usr/bin/env bash
+# Scribe plugin — PostToolUse write-path trigger on Bash (#2901).
+#
+# scribe_prior_art.sh fires before a Write/Edit TOOL CALL. Code written any
+# other way — sed, heredocs, python edit scripts, `cat > file` — never reached
+# it, so a whole class of edits (the ones a long session makes most) got no
+# prior-art hint, no ledger feed and no duplicate-family warning. This hook
+# closes that: after EVERY Bash call it asks git what changed in the working
+# tree since it last looked, and runs the same arms on the definitions that
+# were just written — the local by-name duplicate arm, the recorded prior-art
+# arms and the ledger's derive/divergence checks (#2900/#2793), via the same
+# /api/plugin/prior-art endpoint the pre-write hook uses.
+#
+# Post-hoc by a few seconds, in the same moment and the same session: "the
+# copy just landed; here is its family" — not "an audit found it later".
+#
+# Cheap when nothing changed: one `git status`. State per session, beside the
+# pre-write hook's (its three dedup channels are SHARED, so a family named by
+# one hook is not named again by the other):
+#   ${TMPDIR:-/tmp}/scribe-afterwrite/<sid>.snap   path<TAB>blob-hash of every
+#                                                  dirty/untracked file last seen
+#   ${TMPDIR:-/tmp}/scribe-priorart/<sid>.*        the dedup channels
+#
+# NEVER BLOCKS. It returns `additionalContext` only (no decision — there is
+# nothing left to decide, the write already happened). Any failure —
+# unconfigured, unreachable, not a git repo, malformed — exits 0 in silence.
+#
+# Config (same as the other hooks):
+#   CLAUDE_PLUGIN_OPTION_API_ENDPOINT  base URL, no trailing slash
+#   CLAUDE_PLUGIN_OPTION_API_TOKEN     fmcp_ API key (sensitive)
+#   SCRIBE_URL / SCRIBE_TOKEN override for the settings.json dogfooding path.
+set -uo pipefail
+
+command -v jq  >/dev/null 2>&1 || exit 0
+command -v git >/dev/null 2>&1 || exit 0
+
+# shellcheck source=plugin/hooks/scribe_defs.sh
+. "$(dirname "${BASH_SOURCE[0]}")/scribe_defs.sh"
+
+# PostToolUse delivers { session_id, cwd, tool_name, tool_input, tool_response }.
+event=$(cat 2>/dev/null || true)
+tool_name=$(printf '%s' "$event" | jq -r '.tool_name // empty' 2>/dev/null) || exit 0
+[ "$tool_name" = "Bash" ] || exit 0
+session_id=$(printf '%s' "$event" | jq -r '.session_id // empty' 2>/dev/null) || session_id=""
+event_cwd=$(printf '%s' "$event" | jq -r '.cwd // empty' 2>/dev/null) || event_cwd=""
+work_dir=${event_cwd:-${CLAUDE_PROJECT_DIR:-$PWD}}
+repo_root=$(git -C "$work_dir" rev-parse --show-toplevel 2>/dev/null) || exit 0
+[ -n "$repo_root" ] || exit 0
+
+safe_sid=$(printf '%s' "${session_id:-nosession}" | tr -c 'A-Za-z0-9._-' '_')
+snap_dir="${TMPDIR:-/tmp}/scribe-afterwrite"
+mkdir -p "$snap_dir" 2>/dev/null || true
+snap="$snap_dir/${safe_sid}.snap"
+
+# What is dirty now: every modified / added / untracked path, with the blob
+# hash of its working-tree content. Hash, not mtime: portable (no stat
+# flags), exact (a touch is not a change), and untracked files hash the same
+# way tracked ones do.
+current=""
+while IFS= read -r line; do
+  [ -n "$line" ] || continue
+  status=${line:0:2}
+  path=${line:3}
+  case "$status" in
+    D*|*D) continue ;;                       # a deletion defines nothing
+  esac
+  case "$path" in
+    *" -> "*) path=${path##* -> } ;;          # rename: the new name
+  esac
+  # Porcelain quotes paths with special characters; those are skipped rather
+  # than unquoted badly — a filename needing quotes is not where shapes live.
+  case "$path" in
+    \"*) continue ;;
+  esac
+  [ -f "$repo_root/$path" ] || continue
+  sha=$(git -C "$repo_root" hash-object -- "$path" 2>/dev/null) || continue
+  current="${current}${path}"$'\t'"${sha}"$'\n'
+done < <(git -C "$repo_root" status --porcelain --untracked-files=all 2>/dev/null)
+
+previous=""
+[ -f "$snap" ] && previous=$(cat "$snap" 2>/dev/null || true)
+first_run=0
+[ -f "$snap" ] || first_run=1
+# Write the new snapshot NOW, before anything can fail below — the next call
+# must compare against this tree, whatever happens to this one's hint.
+printf '%s' "$current" > "$snap" 2>/dev/null || true
+
+# Changed = a (path, hash) pair not in the previous snapshot. On the very
+# first call of a session there is no previous snapshot; rather than report
+# every pre-existing dirty file as "just written", take only files touched in
+# the last minute — the Bash call that just ran is the likely author.
+changed=""
+while IFS=$'\t' read -r path sha; do
+  [ -n "${path:-}" ] || continue
+  if [ "$first_run" = 1 ]; then
+    [ -n "$(find "$repo_root/$path" -mmin -1 2>/dev/null)" ] || continue
+  else
+    case "$previous" in
+      *"${path}"$'\t'"${sha}"*) continue ;;
+    esac
+  fi
+  scribe_skip_path "$path" && continue
+  changed="${changed}${path}"$'\n'
+done <<< "$current"
+[ -n "$changed" ] || exit 0
+
+url=${SCRIBE_URL:-${CLAUDE_PLUGIN_OPTION_API_ENDPOINT:-}}
+token=${SCRIBE_TOKEN:-${CLAUDE_PLUGIN_OPTION_API_TOKEN:-}}
+case "$url"   in *'${'*) url="" ;;  esac
+case "$token" in *'${'*) token="" ;; esac
+repo=$(git -C "$repo_root" remote get-url origin 2>/dev/null || true)
+repo_q=""
+if [ -n "$repo" ]; then
+  enc=$(printf '%s' "$repo" | jq -sRr '@uri' 2>/dev/null) || enc=""
+  [ -n "$enc" ] && repo_q="&repo=${enc}"
+fi
+
+# The dedup channels are the PRE-write hook's files, on purpose (see header).
+state_dir="${TMPDIR:-/tmp}/scribe-priorart"
+mkdir -p "$state_dir" 2>/dev/null || true
+idfile="$state_dir/${safe_sid}.ids"
+syncfile="$state_dir/${safe_sid}.sync.ids"
+derivefile="$state_dir/${safe_sid}.derive.ids"
+
+combined=""
+n_files=0
+while IFS= read -r rel_path; do
+  [ -n "${rel_path:-}" ] || continue
+  # A Bash call that rewrote many files is a refactor or a generator, not a
+  # shape being instantiated; four is enough to name what matters.
+  n_files=$((n_files + 1))
+  [ "$n_files" -le 4 ] || break
+  file_path="$repo_root/$rel_path"
+
+  # The code just written: the ADDED lines of the uncommitted diff for a
+  # tracked file (sed, not cut: this strips one marker char per line, it is
+  # not a payload cap), the whole file when untracked.
+  if git -C "$repo_root" ls-files --error-unmatch -- "$rel_path" >/dev/null 2>&1; then
+    code=$(git -C "$repo_root" diff -U0 -- "$rel_path" 2>/dev/null | grep '^+' | grep -v '^+++' | sed 's/^+//') || code=""
+  else
+    code=$(cat "$file_path" 2>/dev/null) || code=""
+  fi
+  [ -n "$code" ] || continue
+  names=$(printf '%s' "$code" | scribe_defs | sort -u | head -12) || names=""
+  # Nothing DEFINED in what was written (prose, data, a call-site edit) →
+  # nothing to say; the arms are about shapes.
+  [ -n "$names" ] || continue
+
+  local_lines=$(scribe_local_dups "$repo_root" "$rel_path" <<< "$names") || local_lines=""
+  local_context=""
+  if [ -n "$local_lines" ]; then
+    local_context="> Already defined elsewhere in this repo — \`${rel_path}\` (just written) adds another copy; check before keeping it (\`git grep\` shown; a nudge, not a gate):"$'\n'"${local_lines}"
+  fi
+
+  context=""
+  body=""
+  if [ -n "$url" ] && [ -n "$token" ]; then
+    q=$(printf '%s' "$code" | head -c 1200)
+    path_enc=$(printf '%s' "$rel_path" | jq -sRr '@uri' 2>/dev/null) || path_enc=""
+    code_enc=$(printf '%s' "$q" | jq -sRr '@uri' 2>/dev/null) || code_enc=""
+    shapes_q=""
+    enc=$(printf '%s\n' "$names" \
+      | awk -F'\t' 'NF>=2 {printf "%s%s:%s", (n++?",":""), $1, $2}' \
+      | jq -sRr '@uri' 2>/dev/null) || enc=""
+    [ -n "$enc" ] && shapes_q="&shapes=${enc}"
+    exclude_q=""; sync_exclude_q=""; derive_exclude_q=""
+    if [ -f "$idfile" ]; then
+      seen=$(tr '\n' ',' < "$idfile" 2>/dev/null | sed 's/,$//')
+      [ -n "$seen" ] && exclude_q="&exclude_ids=${seen}"
+    fi
+    if [ -f "$syncfile" ]; then
+      sync_seen=$(tr '\n' ',' < "$syncfile" 2>/dev/null | sed 's/,$//')
+      [ -n "$sync_seen" ] && sync_exclude_q="&exclude_sync_ids=${sync_seen}"
+    fi
+    if [ -f "$derivefile" ]; then
+      derive_seen=$(tr '\n' ',' < "$derivefile" 2>/dev/null | sed 's/,$//' | jq -sRr '@uri' 2>/dev/null) || derive_seen=""
+      [ -n "$derive_seen" ] && derive_exclude_q="&exclude_derive=${derive_seen}"
+    fi
+    if [ -n "$path_enc" ]; then
+      body=$(curl -fsS --max-time 4 \
+        -H "Authorization: Bearer ${token}" \
+        "${url%/}/api/plugin/prior-art?path=${path_enc}&code=${code_enc}${repo_q}${exclude_q}${sync_exclude_q}${derive_exclude_q}${shapes_q}" 2>/dev/null) || body=""
+    fi
+    if [ -n "$body" ]; then
+      context=$(printf '%s' "$body" | jq -r '.context // empty' 2>/dev/null) || context=""
+      if [ -n "$context" ]; then
+        printf '%s' "$body" | jq -r '((.note_ids // []) - (.sync_note_ids // []))[]?' 2>/dev/null >> "$idfile" || true
+        printf '%s' "$body" | jq -r '(.sync_note_ids // [])[]?' 2>/dev/null >> "$syncfile" || true
+        printf '%s' "$body" | jq -r '(.derive_keys // [])[]?' 2>/dev/null >> "$derivefile" || true
+      fi
+    fi
+  fi
+
+  # The record nudge (#2664), same gate as the pre-write hook: duplication
+  # demonstrated locally AND nothing recorded for it.
+  if [ -n "$local_lines" ]; then
+    n_recorded=$(printf '%s' "$body" | jq -r '.note_ids | length' 2>/dev/null) || n_recorded=0
+    if [ "${n_recorded:-0}" = "0" ] || [ "$n_recorded" = "" ]; then
+      local_context="${local_context}"$'\n'"> None of those existing copies is recorded in Scribe. If the version just written is the canonical one — or this edit is consolidating the copies — record it now with create_snippet so the next session is offered it instead of writing another copy."
+    fi
+  fi
+
+  part="$local_context"
+  if [ -n "$context" ]; then
+    [ -n "$part" ] && part="${part}"$'\n'
+    part="${part}${context}"
+  fi
+  [ -n "$part" ] || continue
+  [ -n "$combined" ] && combined="${combined}"$'\n'
+  combined="${combined}${part}"
+done <<< "$changed"
+
+[ -n "$combined" ] || exit 0
+jq -n --arg c "$combined" \
+  '{hookSpecificOutput: {hookEventName: "PostToolUse", additionalContext: $c}}'
+exit 0

plugin/hooks/scribe_defs.sh

@@ -0,0 +1,109 @@
+#!/usr/bin/env bash
+# shellcheck shell=bash
+# Scribe plugin — the pieces the two write-path hooks share (#2901).
+#
+# scribe_prior_art.sh fires BEFORE a Write/Edit tool call; scribe_after_write.sh
+# fires AFTER a Bash tool call and diffs the working tree, so code written by
+# sed/heredocs/scripts gets the same prior-art and ledger checks. Both need the
+# same three things, kept here so they cannot drift apart:
+#
+#   scribe_skip_path PATH      formats that hold prose or data, not shapes
+#   scribe_defs                stdin code → "kind<TAB>name" per definition
+#   scribe_local_dups ROOT REL "kind<TAB>name" lines on stdin → the by-name
+#                              local-duplicate lines (ARM 1, #2280)
+#
+# Sourced, not executed: `. "$(dirname "${BASH_SOURCE[0]}")/scribe_defs.sh"`.
+
+# Skip formats that hold prose or data rather than reusable code. Purely to
+# avoid a pointless round-trip — the server would return nothing for these
+# anyway. Config formats are NOT skipped: a CI workflow or a compose file is
+# often exactly the thing worth reusing.
+scribe_skip_path() {
+  case "$1" in
+    *.md|*.mdx|*.txt|*.rst|*.json|*.lock|*.log|*.csv|*.tsv|*.svg|*.png|*.jpg|*.jpeg|*.gif|*.ico|*.pdf)
+      return 0 ;;
+  esac
+  return 1
+}
+
+# ---------------------------------------------------------------------------
+# kind<TAB>name for each thing a piece of code DEFINES, in source order. One
+# program, two consumers: the local duplicate arm (every definition in the
+# payload) and the ledger feed (#2791, below: the definitions being written,
+# or the one enclosing an Edit). Rule-for-rule mirrored by the server's
+# services/coverage.py extract_shapes — ledger rows are keyed by what THAT
+# sees, so the two must agree on what counts as a definition.
+scribe_defs() {
+  awk '
+    {
+      # CSS class definition: .name { or .name,
+      if (match($0, /^[[:space:]]*\.[A-Za-z][A-Za-z0-9_-]*[[:space:]]*[,{]/)) {
+        t = $0; sub(/^[[:space:]]*\./, "", t); sub(/[[:space:]]*[,{].*$/, "", t)
+        if (t != "") print "css\t" t; next
+      }
+      line = $0; sub(/^[[:space:]]+/, "", line)
+      # Strip leading declaration modifiers so the definition keyword is the
+      # first word regardless of language (export/pub/private/suspend/...).
+      sub(/^((pub(\([a-z]+\))?|export|default|private|internal|protected|public|static|suspend|async|open|sealed|data|abstract|final|inline|unsafe|extern|override)[[:space:]]+)*/, "", line)
+      # Go method with receiver: func (r *T) Name(
+      if (match(line, /^func[[:space:]]*\([^)]*\)[[:space:]]*[A-Za-z_]/)) {
+        t = line; sub(/^func[[:space:]]*\([^)]*\)[[:space:]]*/, "", t)
+        sub(/[^A-Za-z0-9_].*$/, "", t)
+        if (t != "") print "sym\t" t; next
+      }
+      # Keyword-announced definitions, functions and named types alike.
+      # Dunders are skipped: every class defines __init__, so "already defined
+      # in N other files" is guaranteed noise for them — and noise is what
+      # teaches sessions to skip the hint.
+      if (match(line, /^(function|def|class|func|fun|fn|sub|struct|trait|interface|enum|object|protocol|type)[[:space:]]+[A-Za-z_$]/)) {
+        t = line; sub(/^[a-z]+[[:space:]]+/, "", t)
+        sub(/[^A-Za-z0-9_$].*$/, "", t)
+        if (t != "" && t !~ /^__.*__$/) print "sym\t" t; next
+      }
+      # Arrow/expression assignment: const name = (…) / let name = async (
+      if (match(line, /^(const|let)[[:space:]]+[A-Za-z_$][A-Za-z0-9_$]*[[:space:]]*=[[:space:]]*(async[[:space:]]*)?[(<]/)) {
+        t = line; sub(/^(const|let)[[:space:]]+/, "", t)
+        sub(/[^A-Za-z0-9_$].*$/, "", t)
+        if (t != "") print "sym\t" t; next
+      }
+    }
+  ' 2>/dev/null
+}
+
+
+# ---------------------------------------------------------------------------
+# ARM 1 — BY NAME, LOCALLY (#2280). Does a definition of this already exist?
+#
+# The recorded arms ask Scribe what was RECORDED; the ledger arm (#2900) asks
+# what a BOUND repo's ledger knows. A helper nobody recorded, in a repo nobody
+# bound, is invisible to both — which is how `.btn-primary` came to be defined
+# four times, already diverged. This arm asks the one question only the
+# developer's machine can answer, inside the repo, holding the code about to
+# be written: no index, no storage, no server — it runs even on an install
+# that has never configured Scribe.
+#
+# Definition-shaped patterns only. Grepping for bare occurrences would match
+# every CALL site and drown the real finding — and a hint that is mostly noise
+# is one people learn to skip, which is worse than none. ALL code, not a
+# language shortlist (#2682): the same keyword family scribe_defs announces.
+#
+# $1 repo root, $2 repo-relative path of the file being written (excluded from
+# the grep — it would always match itself on an Edit). Definitions on stdin.
+# Prints one "> - `name` is already defined in N other file(s): …" per hit.
+scribe_local_dups() {
+  local root="$1" rel="$2" kind name pat hits count label files
+  while IFS=$'\t' read -r kind name; do
+    [ -n "${name:-}" ] || continue
+    case "$kind" in
+      css) pat="^[[:space:]]*\.${name}[[:space:]]*[,{]" ;;
+      *)   pat="(function|def|class|func|fun|fn|sub|struct|trait|interface|enum|object|protocol|type)[[:space:]]+${name}[^A-Za-z0-9_]|func[[:space:]]*\([^)]*\)[[:space:]]*${name}[[:space:]]*\(|(const|let)[[:space:]]+${name}[[:space:]]*=" ;;
+    esac
+    # -I skips binaries; :(exclude) drops the file being written.
+    hits=$(git -C "$root" grep -I -l -E -e "$pat" -- . ":(exclude)${rel}" 2>/dev/null | head -4) || hits=""
+    [ -n "$hits" ] || continue
+    count=$(printf '%s\n' "$hits" | grep -c . 2>/dev/null || echo 0)
+    label=$([ "$kind" = css ] && printf '.%s' "$name" || printf '%s' "$name")
+    files=$(printf '%s' "$hits" | tr '\n' ' ' | sed 's/ $//')
+    printf '> - `%s` is already defined in %s other file(s): %s\n' "$label" "$count" "$files"
+  done
+}

plugin/hooks/scribe_prior_art.sh

@@ -51,14 +51,12 @@ code=$(printf '%s' "$event" | jq -r '
   .tool_input.content // .tool_input.file_content //
   .tool_input.new_string // .tool_input.new_str // empty' 2>/dev/null) || code=""
 
-# Skip formats that hold prose or data rather than reusable code. Purely to
+# Shared with the after-write hook (#2901): the prose/data skip list, the
-# avoid a pointless round-trip — the server would return nothing for these
+# definition extractor and the local by-name duplicate arm live in
-# anyway. Config formats are NOT skipped: a CI workflow or a compose file is
+# scribe_defs.sh so the two hooks cannot drift apart.
-# often exactly the thing worth reusing.
+# shellcheck source=plugin/hooks/scribe_defs.sh
-case "$file_path" in
+. "$(dirname "${BASH_SOURCE[0]}")/scribe_defs.sh"
-  *.md|*.mdx|*.txt|*.rst|*.json|*.lock|*.log|*.csv|*.tsv|*.svg|*.png|*.jpg|*.jpeg|*.gif|*.ico|*.pdf)
+scribe_skip_path "$file_path" && exit 0
-    exit 0 ;;
-esac
 
 # Snippet locations are recorded repo-relative, so send a repo-relative path —
 # an absolute one would simply match nothing. Resolved BEFORE the config gate
@@ -73,78 +71,8 @@ if [ -n "$repo_root" ]; then
   esac
 fi
 
-# ---------------------------------------------------------------------------
+# ARM 1 — BY NAME, LOCALLY (#2280): does a definition of this already exist
-# ARM 1 — BY NAME, LOCALLY (#2280). Does a definition of this already exist?
+# in the repo? (scribe_local_dups in scribe_defs.sh carries the why.)
-#
-# The other two arms ask Scribe what was RECORDED. Scribe has never read a line
-# of the codebase, so a helper nobody thought to record is invisible to them —
-# which is how `.btn-primary` came to be defined four times, in four scoped
-# stylesheets, already diverged. It was never a snippet, so no threshold and no
-# query rewrite could ever have surfaced it.
-#
-# This arm closes that by asking the only question the record cannot answer,
-# in the only place that can: the hook already runs on the developer's machine,
-# inside the repo, holding the code about to be written. No index, no storage,
-# no staleness, and no server — it deliberately runs even on an install that
-# has never configured Scribe.
-#
-# Definition-shaped patterns only. Grepping for bare occurrences would match
-# every CALL site and drown the real finding — and a hint that is mostly noise
-# is one people learn to skip, which is worse than none.
-#
-# ALL code, not a language shortlist (#2682): the detector was born covering
-# only the languages of the repo it was written in, which silently amputated
-# this whole arm — and the record nudge gated on it — for every Go/Kotlin/Rust
-# project. Definitions are announced by a small keyword family across
-# languages (func/fun/fn/function/def/sub · class/struct/trait/interface/
-# enum/object/protocol/type), so one modifier-strip + keyword match covers
-# them all. Known out of scope: keyword-less declaration syntax (C/Java/Dart
-# `ReturnType name(...)`) needs a real parser, and `impl` blocks are excluded
-# because several per type is normal Rust, not duplication.
-# ---------------------------------------------------------------------------
-# kind<TAB>name for each thing a piece of code DEFINES, in source order. One
-# program, two consumers: the local duplicate arm (every definition in the
-# payload) and the ledger feed (#2791, below: the definitions being written,
-# or the one enclosing an Edit). Rule-for-rule mirrored by the server's
-# services/coverage.py extract_shapes — ledger rows are keyed by what THAT
-# sees, so the two must agree on what counts as a definition.
-scribe_defs() {
-  awk '
-    {
-      # CSS class definition: .name { or .name,
-      if (match($0, /^[[:space:]]*\.[A-Za-z][A-Za-z0-9_-]*[[:space:]]*[,{]/)) {
-        t = $0; sub(/^[[:space:]]*\./, "", t); sub(/[[:space:]]*[,{].*$/, "", t)
-        if (t != "") print "css\t" t; next
-      }
-      line = $0; sub(/^[[:space:]]+/, "", line)
-      # Strip leading declaration modifiers so the definition keyword is the
-      # first word regardless of language (export/pub/private/suspend/...).
-      sub(/^((pub(\([a-z]+\))?|export|default|private|internal|protected|public|static|suspend|async|open|sealed|data|abstract|final|inline|unsafe|extern|override)[[:space:]]+)*/, "", line)
-      # Go method with receiver: func (r *T) Name(
-      if (match(line, /^func[[:space:]]*\([^)]*\)[[:space:]]*[A-Za-z_]/)) {
-        t = line; sub(/^func[[:space:]]*\([^)]*\)[[:space:]]*/, "", t)
-        sub(/[^A-Za-z0-9_].*$/, "", t)
-        if (t != "") print "sym\t" t; next
-      }
-      # Keyword-announced definitions, functions and named types alike.
-      # Dunders are skipped: every class defines __init__, so "already defined
-      # in N other files" is guaranteed noise for them — and noise is what
-      # teaches sessions to skip the hint.
-      if (match(line, /^(function|def|class|func|fun|fn|sub|struct|trait|interface|enum|object|protocol|type)[[:space:]]+[A-Za-z_$]/)) {
-        t = line; sub(/^[a-z]+[[:space:]]+/, "", t)
-        sub(/[^A-Za-z0-9_$].*$/, "", t)
-        if (t != "" && t !~ /^__.*__$/) print "sym\t" t; next
-      }
-      # Arrow/expression assignment: const name = (…) / let name = async (
-      if (match(line, /^(const|let)[[:space:]]+[A-Za-z_$][A-Za-z0-9_$]*[[:space:]]*=[[:space:]]*(async[[:space:]]*)?[(<]/)) {
-        t = line; sub(/^(const|let)[[:space:]]+/, "", t)
-        sub(/[^A-Za-z0-9_$].*$/, "", t)
-        if (t != "") print "sym\t" t; next
-      }
-    }
-  ' 2>/dev/null
-}
-
 names=""
 if [ -n "$code" ]; then
   names=$(printf '%s' "$code" | scribe_defs | sort -u | head -12) || names=""
@@ -152,21 +80,8 @@ fi
 
 local_lines=""
 if [ -n "$repo_root" ] && [ -n "$names" ]; then
-  while IFS=$'\t' read -r kind name; do
+  local_lines=$(scribe_local_dups "$repo_root" "$rel_path" <<< "$names") || local_lines=""
-    [ -n "${name:-}" ] || continue
+  [ -n "$local_lines" ] && local_lines="${local_lines}"$'\n'
-    case "$kind" in
-      css) pat="^[[:space:]]*\.${name}[[:space:]]*[,{]" ;;
-      *)   pat="(function|def|class|func|fun|fn|sub|struct|trait|interface|enum|object|protocol|type)[[:space:]]+${name}[^A-Za-z0-9_]|func[[:space:]]*\([^)]*\)[[:space:]]*${name}[[:space:]]*\(|(const|let)[[:space:]]+${name}[[:space:]]*=" ;;
-    esac
-    # -I skips binaries; :(exclude) drops the file being written, which would
-    # otherwise always match itself on an Edit.
-    hits=$(git -C "$repo_root" grep -I -l -E -e "$pat" -- . ":(exclude)${rel_path}" 2>/dev/null | head -4) || hits=""
-    [ -n "$hits" ] || continue
-    count=$(printf '%s\n' "$hits" | grep -c . 2>/dev/null || echo 0)
-    label=$([ "$kind" = css ] && printf '.%s' "$name" || printf '%s' "$name")
-    files=$(printf '%s' "$hits" | tr '\n' ' ' | sed 's/ $//')
-    local_lines="${local_lines}> - \`${label}\` is already defined in ${count} other file(s): ${files}"$'\n'
-  done <<< "$names"
 fi
 
 local_context=""

scripts/check_plugin.py

@@ -203,6 +203,17 @@ SMOKE_EVENTS: dict[str, str] = {
     ),
     "scribe_sync_processes.sh": json.dumps({"source": "startup"}),
     "scribe_session_context.sh": json.dumps({"source": "startup"}),
+    # The after-write hook (#2901) diffs the working tree; on CI's clean
+    # checkout there is nothing to report, so silence is the right assertion.
+    # (On a dirty local tree with a definition just written it may speak —
+    # that is the hook working, not a failure of the contract.)
+    "scribe_after_write.sh": json.dumps(
+        {"session_id": "smoke", "cwd": ".", "tool_name": "Bash",
+         "tool_input": {"command": "true"}, "tool_response": {}}
+    ),
+    # The shared library is sourced, never run; executed bare it defines
+    # functions and exits — silent by construction.
+    "scribe_defs.sh": "",
 }
 
 # The one hook that legitimately produces output with no credentials.

tests/test_after_write_hook.py

@@ -0,0 +1,145 @@
+"""The PostToolUse after-write hook (#2901): code written through Bash — sed,
+heredocs, scripts — gets the same prior-art / ledger checks as a Write/Edit.
+
+Runs the real shell against a temp git repo and a throwaway HTTP sink, like
+the pre-write hook's end-to-end tests. Skips where the hook's tools are
+missing; asserts on content where they are present."""
+from __future__ import annotations
+
+import http.server
+import json
+import os
+import shutil
+import subprocess
+import threading
+import urllib.parse
+from pathlib import Path
+
+import pytest
+
+PLUGIN = Path(__file__).resolve().parents[1] / "plugin"
+HOOK = PLUGIN / "hooks" / "scribe_after_write.sh"
+
+
+def _env(tmp_path, url="http://127.0.0.1:9"):
+    for tool in ("git", "jq", "curl", "bash"):
+        if shutil.which(tool) is None:
+            pytest.skip(f"hook runtime tool {tool!r} not installed")
+    return {"PATH": os.environ["PATH"], "SCRIBE_URL": url, "SCRIBE_TOKEN": "t",
+            "TMPDIR": str(tmp_path), "HOME": str(tmp_path),
+            "GIT_AUTHOR_NAME": "t", "GIT_AUTHOR_EMAIL": "t@x", 
+            "GIT_COMMITTER_NAME": "t", "GIT_COMMITTER_EMAIL": "t@x"}
+
+
+def _repo(tmp_path, env):
+    repo = tmp_path / "repo"
+    repo.mkdir()
+    subprocess.run(["git", "init", "-q"], cwd=repo, check=True, env=env)
+    (repo / "b.py").write_text("def one():\n    return 1\n")
+    (repo / "c.py").write_text("def slug(t):\n    return t.lower()\n")
+    subprocess.run(["git", "add", "."], cwd=repo, check=True, env=env)
+    subprocess.run(["git", "commit", "-q", "-m", "base"], cwd=repo, check=True, env=env)
+    return repo
+
+
+def _run(repo, env, session="s-after-1", tool="Bash"):
+    out = subprocess.run(
+        ["bash", str(HOOK)],
+        input=json.dumps({"session_id": session, "cwd": str(repo), "tool_name": tool,
+                          "tool_input": {"command": "cat > x"}, "tool_response": {}}),
+        capture_output=True, text=True, env=env,
+    )
+    assert out.returncode == 0, out.stderr
+    return out.stdout
+
+
+class _Sink(http.server.BaseHTTPRequestHandler):
+    seen: list[dict] = []
+    reply = b'{"context":"> family named","note_ids":[],"sync_note_ids":[],"derive_keys":["dup:483a"]}'
+
+    def do_GET(self):
+        type(self).seen.append(urllib.parse.parse_qs(urllib.parse.urlparse(self.path).query))
+        self.send_response(200)
+        self.send_header("Content-Type", "application/json")
+        self.end_headers()
+        self.wfile.write(type(self).reply)
+
+    def log_message(self, *a):
+        pass
+
+
+@pytest.fixture
+def sink():
+    _Sink.seen = []
+    server = http.server.HTTPServer(("127.0.0.1", 0), _Sink)
+    threading.Thread(target=server.serve_forever, daemon=True).start()
+    try:
+        yield server
+    finally:
+        server.shutdown()
+
+
+def test_after_write_names_what_bash_just_wrote_then_stays_quiet_until_the_next_change(tmp_path, sink):
+    env = _env(tmp_path, url=f"http://127.0.0.1:{sink.server_port}")
+    repo = _repo(tmp_path, env)
+    # "A Bash call" wrote an untracked stylesheet and appended to a tracked file.
+    (repo / "a.css").write_text(".log-empty {\n  color: red;\n}\n")
+    (repo / "b.py").write_text("def one():\n    return 1\n\ndef slug(t):\n    return t\n")
+    out = _run(repo, env)
+    by_path = {q["path"][0]: q for q in _Sink.seen}
+    assert set(by_path) == {"a.css", "b.py"}           # repo-relative, like the pre hook
+    assert by_path["a.css"]["shapes"] == ["css:log-empty"]
+    assert by_path["b.py"]["shapes"] == ["sym:slug"]
+    # Added lines only for the tracked file — the existing def is not "just written".
+    assert "def slug" in by_path["b.py"]["code"][0] and "def one" not in by_path["b.py"]["code"][0]
+    ctx = json.loads(out)["hookSpecificOutput"]
+    assert ctx["hookEventName"] == "PostToolUse"
+    assert "> family named" in ctx["additionalContext"]
+    # The local by-name arm rides along: `slug` already lives in c.py.
+    assert "`slug` is already defined in 1 other file(s): c.py" in ctx["additionalContext"]
+    # Derive keys landed on the SHARED channel the pre-write hook reads.
+    state = tmp_path / "scribe-priorart" / "s-after-1.derive.ids"
+    assert "dup:483a" in state.read_text().split()
+
+    # Nothing changed → one git status, no request, no output.
+    _Sink.seen = []
+    assert _run(repo, env) == ""
+    assert _Sink.seen == []
+
+    # Another change → only that file, and the dedup channel goes back up.
+    (repo / "a.css").write_text(".log-empty {\n  color: red;\n}\n.other {\n  margin: 0;\n}\n")
+    _run(repo, env)
+    assert [q["path"][0] for q in _Sink.seen] == ["a.css"]
+    assert _Sink.seen[0]["exclude_derive"] == ["dup:483a"]
+    assert set(_Sink.seen[0]["shapes"][0].split(",")) == {"css:log-empty", "css:other"}
+
+
+def test_after_write_is_silent_where_it_has_nothing_to_say(tmp_path):
+    env = _env(tmp_path)
+    repo = _repo(tmp_path, env)
+    # Not a Bash call → nothing (hooks.json matches Bash, the script re-checks).
+    (repo / "a.css").write_text(".x {\n  color: red;\n}\n")
+    assert _run(repo, env, tool="Write") == ""
+    # Not a git repo → nothing.
+    loose = tmp_path / "loose"
+    loose.mkdir()
+    (loose / "a.css").write_text(".x {\n  color: red;\n}\n")
+    assert _run(loose, env, session="s-loose") == ""
+    # A change that defines nothing (prose, a call-site edit) → nothing, even
+    # with the server unreachable (port 9 refuses): no definitions, no arms.
+    (repo / "README.md").write_text("# notes\n")
+    (repo / "b.py").write_text("def one():\n    return one_more()\n")
+    assert _run(repo, env, session="s-quiet") == ""
+
+
+def test_after_write_local_arm_and_record_nudge_work_without_a_server(tmp_path):
+    """The local by-name arm needs no instance (#2280) and the record nudge
+    (#2664) fails open with it — a refused connection stands in for the
+    instance."""
+    env = _env(tmp_path)
+    repo = _repo(tmp_path, env)
+    (repo / "d.py").write_text("def slug(t):\n    return t.lower()\n")
+    out = _run(repo, env, session="s-local")
+    ctx = json.loads(out)["hookSpecificOutput"]["additionalContext"]
+    assert "`slug` is already defined in 1 other file(s): c.py" in ctx
+    assert "create_snippet" in ctx