@mmerterden/multi-agent-pipeline 20.3.0 → 20.4.0

package/pipeline/multi-agent-refs/channels/pr.md

@@ -251,7 +251,7 @@ The two adapters run in the same phase over the same source markdown, so the fai
 ## Reviewer-preserving Bitbucket payload (required)
 
 ```bash
-PR_JSON=$(curl -s -u "$BB_USER:$BB_TOKEN" "$PR_URL")
+PR_JSON=$(curl -s --config <(printf 'user = "%s:%s"\n' "$BB_USER" "$BB_TOKEN") "$PR_URL")
 REVIEWERS=$(jq '[.reviewers[] | {user: {name: .user.name}, approved, status}]' <<< "$PR_JSON")
 VERSION=$(jq -r '.version' <<< "$PR_JSON")
 TITLE=$(jq -r '.title' <<< "$PR_JSON")
@@ -268,7 +268,7 @@ jq -n --rawfile body /tmp/channels-$TASK_ID-pr.md \
         reviewers: $reviewers, fromRef: $fromRef, toRef: $toRef, draft: $draft}' \
   > /tmp/channels-$TASK_ID-pr-payload.json
 
-curl -s -X PUT -u "$BB_USER:$BB_TOKEN" -H "Content-Type: application/json" \
+curl -s -X PUT --config <(printf 'user = "%s:%s"\n' "$BB_USER" "$BB_TOKEN") -H "Content-Type: application/json" \
      --data-binary @/tmp/channels-$TASK_ID-pr-payload.json "$PR_URL"
 ```
 

package/pipeline/multi-agent-refs/features/doctor.md

@@ -167,7 +167,8 @@ reaches Phase 4 and stops there.
 ### hook-coverage
 
 The blocking `PreToolUse` gates from `templates/claude-hooks.json` are present in
-`settings.json`. WARN. SKIP when the template is not installed - which was the
+`settings.json`, and none of them is registered twice on the same matcher
+(Claude Code runs a duplicate on every matching call). WARN. SKIP when the template is not installed - which was the
 permanent state until the installer began copying `templates/`, and is exactly
 the case this severity exists to make visible.
 

package/pipeline/multi-agent-refs/features/prior-fix-detection.md

@@ -19,7 +19,7 @@ Before any analysis, check if this issue was already fixed by someone else.
 3. **Jira comments/status**  -  fetch latest comments to see if someone reported a fix:
 
    ```bash
-   curl -s -H "Authorization: Bearer $JIRA_TOKEN" \
+   printf 'Authorization: Bearer %s\n' "$JIRA_TOKEN" | curl -s -H @- \
      "$JIRA_BASE/issue/{jiraId}?fields=status,comment&expand=changelog" \
      | jq '.fields.status.name, .fields.comment.comments[-3:][]?.body'
    ```

package/pipeline/multi-agent-refs/generate-issue.md

@@ -79,7 +79,7 @@ Fallback chain (same as `$HOME/.claude/multi-agent-refs/issue-jira-triad.md` Cre
 2. `prefs.global.defaultJiraKey`
 3. Fetch and ask:
 ```bash
-curl -s -H "Authorization: Bearer $TOKEN" \
+printf 'Authorization: Bearer %s\n' "$TOKEN" | curl -s -H @- \
   "https://$ACCOUNT_JIRA_HOST/rest/api/2/project" \
   | python3 -c 'import json,sys; [print(p["key"]+"\t"+p["name"]) for p in json.load(sys.stdin)]'
 ```
@@ -108,7 +108,7 @@ Three GET groups. Run them before drafting anything.
 **5a. Sample recent same-type issues:**
 ```bash
 JQL="project = ${PROJECT_KEY} AND issuetype = ${ISSUE_TYPE} ORDER BY created DESC"
-curl -s -H "Authorization: Bearer $TOKEN" \
+printf 'Authorization: Bearer %s\n' "$TOKEN" | curl -s -H @- \
   "https://$ACCOUNT_JIRA_HOST/rest/api/2/search?jql=$(printf '%s' "$JQL" | python3 -c 'import sys,urllib.parse; print(urllib.parse.quote(sys.stdin.read()))')&fields=summary,description,labels,components,priority,fixVersions,issuelinks&maxResults=30"
 ```
 Extract from the sample (thresholds are guidance, not hard gates):
@@ -123,7 +123,7 @@ Extract from the sample (thresholds are guidance, not hard gates):
 
 **5b. Field discovery (createmeta):**
 ```bash
-curl -s -H "Authorization: Bearer $TOKEN" \
+printf 'Authorization: Bearer %s\n' "$TOKEN" | curl -s -H @- \
   "https://$ACCOUNT_JIRA_HOST/rest/api/2/issue/createmeta?projectKeys=${PROJECT_KEY}&issuetypeNames=${ISSUE_TYPE}&expand=projects.issuetypes.fields"
 ```
 - 404 or empty `projects[]` (newer Jira DC deprecates the classic form) → fall back to the paged endpoint: `GET /rest/api/2/issue/createmeta/{PROJECT_KEY}/issuetypes` then `GET /rest/api/2/issue/createmeta/{PROJECT_KEY}/issuetypes/{issuetypeId}`.
@@ -136,9 +136,9 @@ curl -s -H "Authorization: Bearer $TOKEN" \
 
 **5c. Board + active sprint (Agile REST):**
 ```bash
-curl -s -H "Authorization: Bearer $TOKEN" \
+printf 'Authorization: Bearer %s\n' "$TOKEN" | curl -s -H @- \
   "https://$ACCOUNT_JIRA_HOST/rest/agile/1.0/board?projectKeyOrId=${PROJECT_KEY}"
-curl -s -H "Authorization: Bearer $TOKEN" \
+printf 'Authorization: Bearer %s\n' "$TOKEN" | curl -s -H @- \
   "https://$ACCOUNT_JIRA_HOST/rest/agile/1.0/board/${BOARD_ID}/sprint?state=active"
 ```
 - Multiple boards → AskUserQuestion (board name + type). One board → silent. Zero boards or Agile API 404 → the sprint option is silently omitted; the preview notes "backlog only (no board found)".
@@ -217,7 +217,7 @@ jq -n --rawfile desc /tmp/generate-issue-$$-escaped.txt \
   '{fields: {project: {key: $key}, issuetype: {name: $type}, summary: $summary, description: $desc}}' \
   > /tmp/generate-issue-$$-payload.json
 # merge approved optional fields (priority / labels / components / epic / assignee / required customs) into .fields with jq before POST
-curl -sS -X POST -H "Authorization: Bearer $TOKEN" -H "Content-Type: application/json" \
+printf 'Authorization: Bearer %s\n' "$TOKEN" | curl -sS -X POST -H @- -H "Content-Type: application/json" \
   --data-binary @/tmp/generate-issue-$$-payload.json \
   "https://$ACCOUNT_JIRA_HOST/rest/api/2/issue"
 ```
@@ -231,20 +231,20 @@ Each step is failure-isolated: on error, warn + continue  -  the issue already e
 1. **Sprint** (when chosen in step 8/9):
 ```bash
 jq -n --arg key "$NEW_KEY" '{issues: [$key]}' > /tmp/generate-issue-$$-sprint.json
-curl -sS -X POST -H "Authorization: Bearer $TOKEN" -H "Content-Type: application/json" \
+printf 'Authorization: Bearer %s\n' "$TOKEN" | curl -sS -X POST -H @- -H "Content-Type: application/json" \
   --data-binary @/tmp/generate-issue-$$-sprint.json \
   "https://$ACCOUNT_JIRA_HOST/rest/agile/1.0/sprint/${SPRINT_ID}/issue"
 ```
    Post-create move instead of a Sprint custom field in the create payload: field-id-agnostic, works on every Jira Software instance.
 2. **Figma remote link** (when `FIGMA_URL` set):
 ```bash
-curl -sS -X POST -H "Authorization: Bearer $TOKEN" -H "Content-Type: application/json" \
+printf 'Authorization: Bearer %s\n' "$TOKEN" | curl -sS -X POST -H @- -H "Content-Type: application/json" \
   -d '{"object":{"url":"'"$FIGMA_URL"'","title":"Figma design"}}' \
   "https://$ACCOUNT_JIRA_HOST/rest/api/2/issue/${NEW_KEY}/remotelink"
 ```
 3. **Swagger remote link** (when `SWAGGER_URL` set):
 ```bash
-curl -sS -X POST -H "Authorization: Bearer $TOKEN" -H "Content-Type: application/json" \
+printf 'Authorization: Bearer %s\n' "$TOKEN" | curl -sS -X POST -H @- -H "Content-Type: application/json" \
   -d '{"object":{"url":"'"$SWAGGER_URL"'","title":"API contract"}}' \
   "https://$ACCOUNT_JIRA_HOST/rest/api/2/issue/${NEW_KEY}/remotelink"
 ```

package/pipeline/multi-agent-refs/outside-the-pipeline.md

@@ -24,10 +24,10 @@ mapping would have given you.
 
 ## Keeping the value out of everything
 
-`curl` config on stdin, so the token never reaches argv (where `ps` can read it):
+The header goes to `curl` on stdin, so the token never reaches argv (where `ps` can read it):
 
 ```bash
-curl --config <(printf 'header = "Authorization: Bearer %s"\n' "$TOKEN") \
+printf 'Authorization: Bearer %s\n' "$TOKEN" | curl -H @- \
      "https://$(jq -r '.global.hosts.jira' "$PREFS")/rest/api/3/issue/PROJ-1"
 ```
 

package/pipeline/multi-agent-refs/phases/phase-0-init.md

@@ -234,7 +234,7 @@ Classify and fetch external data:
    [ -z "$JIRA_KEY" ] && JIRA_KEY="${USER}_Jira_Access_Token"
    # unattended: skipped - a run reads no credential itself (features/unattended-security.md)
    JIRA_TOKEN=$(~/.claude/lib/credential-store.sh get "$JIRA_KEY" 2>/dev/null)
-   curl -s -H "Authorization: Bearer $JIRA_TOKEN" \
+   printf 'Authorization: Bearer %s\n' "$JIRA_TOKEN" | curl -s -H @- \
      "https://{JIRA_HOST}/rest/api/2/issue/{jiraId}?fields=summary,issuetype,status"
    ```
 2. Extract: summary, issueType (Bug/Story/Task/Feature)

package/pipeline/multi-agent-refs/phases/phase-2-dev.md

@@ -177,8 +177,9 @@ For each task (respecting dependency order):
 6. **Intermediate commit** (after each completed task in the plan):
    ```bash
    git -C "{worktreePath}" add -A
-   git -C "{worktreePath}" commit -m "wip({scope}): {task summary} [{jiraId}]"
+   git -C "{worktreePath}" commit -F "/tmp/wip-message-$TASK_ID.txt"
    ```
+   Write `wip({scope}): {task summary} [{jiraId}]` to `/tmp/wip-message-$TASK_ID.txt` with the file tool first. The summary comes from the issue, and inside a double-quoted `-m` the shell would run any `$(...)` in it.
    WIP commits are squashed in Phase 4 before PR. This prevents work loss if a later task fails or the session crashes.
 7. Update task status: `completed` (apply the same verify rule  -  re-read the Status field and retry once on mismatch)
 8. Log: "Phase 2: {task}  -  Test + Build passed"

package/pipeline/multi-agent-refs/phases/phase-4-commit.md

@@ -127,7 +127,7 @@ Branch **deterministically**, no implicit fallback. Read `agent-state.json` and
 
    The script salvages before removing, keeps the branch, and does **not** check it out  -  the user's HEAD and uncommitted work stay put. On success it stamps `worktreeRemovedAt`, `artifactsPath` and `worktreePath: null` into the **salvaged** `agent-state.json` itself. Do NOT write those via `write-state.mjs "$STATE_FILE"`: in single-repo mode that path went away with the worktree, so the write fails, the fields land nowhere, and Phase 5 reads a dead path and silently skips the triage ingest. After a removal re-point `STATE_FILE` at `$(jq -r .artifactsPath <<< "$FIN")/agent-state.json`. Exit 3 is a safe skip (uncommitted changes, unpushed or detached HEAD, `--local`, cwd inside the tree, preference off): report and continue. Never `--force`. Contract: [`features/worktree-finalize.md`]($HOME/.claude/multi-agent-refs/features/worktree-finalize.md).
 
-10. **Issue body update** (GitHub Issue only): if the issue body has a `### Pull Requests` section and/or a `### Progress` table, fill in the PR URL(s) (one row per submodule  -  e.g. `- **common:** {url}`, `- **uicomponents:** {url}`) and flip the Implementation / Testing / Code Connect flags from Pending to Done using whatever marker the template uses (match in place, do NOT introduce new markers). Apply with `gh issue edit {issueNo} --body "{updated body}"` and preserve every other section unchanged.
+10. **Issue body update** (GitHub Issue only): if the issue body has a `### Pull Requests` section and/or a `### Progress` table, fill in the PR URL(s) (one row per submodule  -  e.g. `- **common:** {url}`, `- **uicomponents:** {url}`) and flip the Implementation / Testing / Code Connect flags from Pending to Done using whatever marker the template uses (match in place, do NOT introduce new markers). Apply it with `gh issue edit {issueNo} --body-file <file>`, preserving every other section unchanged.
 11. **NEVER close or resolve the issue**  -  neither GitHub Issue nor Jira. Issues require team review (4 approvals) before closing. Only post a comment with commit/PR URLs.
 12. Log: "Phase 4: Commit {sha}  -  PR #{number}, worktree {removed|kept: <reason>}"
 
@@ -174,13 +174,13 @@ Inputs are read from state and the worktree before writing, not recalled from th
 
 **Default Reviewers**: use `prefs.projects[{project}].defaultReviewers` when set; otherwise ask once ("Default reviewers for this project? comma-separated, or skip") and persist the answer there. Per-remote resolution rules are in "Default reviewers (required on every PR)" below  -  this is the prefs half only.
 
-**Standard PR (bug fix, feature, refactor)**  -  invocation shape; body comes from a file so newlines survive:
+**Standard PR (bug fix, feature, refactor)**  -  invocation shape. Title (`{type}({scope}): {summary} [{jiraId}]`) and body go to files first, so issue text never sits in a quoted argument:
 
 ```bash
-REVIEWERS=$(jq -r '.projects["'$PROJECT'"].defaultReviewers // [] | map("--reviewer " + .) | join(" ")' "$PREFS_FILE")
+REVIEWERS=$(jq -r --arg p "$PROJECT" '.projects[$p].defaultReviewers // [] | map("--reviewer " + .) | join(" ")' "$PREFS_FILE")
 # unattended: skipped - pr-request.mjs replaces the PR (features/unattended-security.md)
 gh pr create --base {baseBranch} --head {branch} $REVIEWERS \
-  --title "{type}({scope}): {summary} [{jiraId}]" \
+  --title "$(cat /tmp/pr-title-$TASK_ID.txt)" \
   --body-file /tmp/pr-body-$TASK_ID.md
 ```
 

package/pipeline/multi-agent-refs/setup/repo-discovery.md

@@ -39,7 +39,7 @@ BB_HOST="$(jq -r '.global.bitbucketHost // empty' "$PREFS_FILE")"
 BB_USER="$(~/.claude/lib/credential-store.sh get "$(jq -r '.global.keychainMapping.bitbucket_user' "$PREFS_FILE")" 2>/dev/null)"
 BB_TOK="$(~/.claude/lib/credential-store.sh get "$(jq -r '.global.keychainMapping.bitbucket_token' "$PREFS_FILE")" 2>/dev/null)"
 CUTOFF_MS=$(( ( $(date +%s) - 7776000 ) * 1000 ))   # 90 days
-curl -sfS -u "$BB_USER:$BB_TOK" \
+curl -sfS --config <(printf 'user = "%s:%s"\n' "$BB_USER" "$BB_TOK") \
   "https://${BB_HOST}/rest/api/1.0/repos?permission=REPO_WRITE&limit=200" \
   | jq -r --argjson cutoff "$CUTOFF_MS" '
       .values[] | select((.updatedDate // 0) >= $cutoff) |
@@ -59,7 +59,7 @@ CUTOFF_ISO="$(date -u -v-90d +%Y-%m-%dT%H:%M:%SZ 2>/dev/null || date -u -d '90 d
 gh api --paginate 'user/repos?affiliation=owner,collaborator&sort=pushed&per_page=100' \
   -q ".[] | select(.pushed_at > \"$CUTOFF_ISO\") | {full: .full_name, clone: .clone_url}"
 # Fallback without gh:
-curl -sfS -H "Authorization: Bearer $GH_TOK" \
+printf 'Authorization: Bearer %s\n' "$GH_TOK" | curl -sfS -H @- \
   'https://api.github.com/user/repos?affiliation=owner,collaborator&sort=pushed&per_page=100'
 ```
 

package/pipeline/schemas/prefs.schema.json

@@ -1820,7 +1820,7 @@
               },
               "upstreamLocalClone": {
                 "type": "string",
-                "description": "Working copy of the upstream repository, checked before the plugin cache. The cache under ~/.claude/plugins/cache is only as fresh as the last `claude marketplace update`, so it is a mirror and never the authority. A private upstream may also be unreachable over the API from the active account, in which case this clone is the only source."
+                "description": "Working copy of the upstream repository, checked before the plugin cache. The cache under ~/.claude/plugins/cache is only as fresh as the last `claude plugin marketplace update`, so it is a mirror and never the authority. A private upstream may also be unreachable over the API from the active account, in which case this clone is the only source."
               },
               "upstreamRepoUrl": {
                 "type": "string",

package/pipeline/scripts/capture-resume.sh

@@ -33,16 +33,9 @@ LOGS="${LOGS_ROOT:-$HOME/.claude/logs/multi-agent}"
 PIPELINE_MEM="$HOME/.claude/memory/multi-agent/_pipeline"
 REVIEW_DAYS=7
 
-# Three buckets, not one line. Measured on the development machine: 22 runs read
-# `in_progress` and every one was over a day old, but they are not one failure.
-# 13 stopped at Phase 0, which is almost entirely questions - those runs never
-# started. 3 had their PR already open and were waiting at Phase 4/7, where the
-# pipeline pauses ON PURPOSE for channel selection. 4 died mid-development.
-#
-# Reporting the newest one of those as "stopped at Phase N" told the truth about
-# one run and hid twenty-one, and it read the same whether the pipeline was
-# waiting for the user or had crashed. A finished job and a dead one should not
-# share a sentence.
+# Unfinished runs are sorted into three buckets because each needs a different
+# action: waiting on the user (an open PR, or a pause for the channel answer),
+# left at a Phase 0 question (nothing built), and stopped mid-development.
 AWAITING_N=0; AWAITING_TASK=""; AWAITING_PHASE=""
 DIED_N=0;     DIED_TASK="";     DIED_PHASE=""
 UNSTARTED_N=0; UNSTARTED_TASK=""
@@ -50,34 +43,35 @@ STALE_TASK=""
 STALE_PHASE=""
 
 if [ -d "$LOGS" ] && command -v jq >/dev/null 2>&1; then
-  # Newest first, so the first hit in each bucket is the one worth naming.
-  # Both layouts are walked: state is written flat as <task-id>/ by
+  # Oldest first, so the first hit in each bucket is the one the banner calls
+  # the oldest. Both layouts are walked: state is written flat as <task-id>/ by
   # phase-tracker.sh and nested as <project>/<task-id>/ by Phase 0, and a
   # reader that knows only one of them under-reports by the size of the other.
+  # artifacts/ holds the copy gc-abandoned takes before it marks a run failed,
+  # so it keeps the status the run had while it was alive.
   while IFS= read -r f; do
     [ -f "$f" ] || continue
     STATUS=$(jq -r '.status // ""' "$f" 2>/dev/null)
-    # Only runs that CLAIM to be running are bucketed. A file with no status is
-    # unknown - eight of them here, six with no currentPhase either - and
-    # reporting unknown as "stopped mid-development" is the same false claim
-    # this change exists to remove, just pointed the other way. If that hides a
-    # real run, the fix is for its writer to record a status.
+    # Only runs that claim to be running are bucketed. A file with no status is
+    # unknown, and reporting unknown as "stopped mid-development" would be a
+    # false claim. The fix for a hidden run is for its writer to record a status.
     # `complete` is the schema spelling and `completed` is in the wild too, so
     # both are excluded; a reader that knows one calls finished runs stale.
     case "$STATUS" in in_progress | paused | awaiting_input) ;; *) continue ;; esac
-    PHASE=$(jq -r '.currentPhase // "?"' "$f" 2>/dev/null)
+    # `phase` is the field name a state written before currentPhase carries.
+    PHASE=$(jq -r '.currentPhase // .phase // "?" | tostring' "$f" 2>/dev/null)
     TASK=$(jq -r '.taskId // .jiraId // ""' "$f" 2>/dev/null)
     [ -n "$TASK" ] || continue
-    # `.pr` is an object in the schema and a bare URL string in three state files
-    # on disk. `jq '.pr.url'` on a string errors, and with stderr suppressed that
-    # reads as "no PR" - so a run whose PR is already open would be classed as dead
-    # and reaped. Read the shape, do not assume it.
+    # `.pr` is an object in the schema and a bare URL string in older state
+    # files. `jq '.pr.url'` on a string errors, and with stderr suppressed that
+    # reads as "no PR", so the shape is read rather than assumed.
     PR=$(jq -r 'if (.pr|type)=="string" then .pr elif (.pr|type)=="object" then (.pr.url // .pr.number // "") else "" end | tostring' "$f" 2>/dev/null)
 
     # Waiting for you, not broken: an open PR means the work landed, and
-    # Phase 5 pauses for channel selection by design (modes.md).
+    # Phase 5 pauses for channel selection by design (modes.md). 6 and 7 are
+    # the same pause in a state written before the six-phase numbering.
     if [ "$STATUS" = "awaiting_input" ] || [ -n "$PR" ] ||
-      [ "$PHASE" = "6" ] || [ "$PHASE" = "7" ]; then
+      [ "$PHASE" = "5" ] || [ "$PHASE" = "6" ] || [ "$PHASE" = "7" ]; then
       AWAITING_N=$((AWAITING_N + 1))
       [ -z "$AWAITING_TASK" ] && { AWAITING_TASK="$TASK"; AWAITING_PHASE="$PHASE"; }
     elif [ "$PHASE" = "0" ]; then
@@ -90,7 +84,7 @@ if [ -d "$LOGS" ] && command -v jq >/dev/null 2>&1; then
       [ -z "$DIED_TASK" ] && { DIED_TASK="$TASK"; DIED_PHASE="$PHASE"; }
     fi
   done <<EOF
-$(find "$LOGS" -name agent-state.json -type f -maxdepth 4 -print0 2>/dev/null | xargs -0 ls -t 2>/dev/null)
+$(find "$LOGS" -maxdepth 4 -name agent-state.json -type f -not -path '*/artifacts/*' -print0 2>/dev/null | xargs -0 ls -tr 2>/dev/null)
 EOF
   # Kept for the JSON consumers that read staleTask/stalePhase.
   if [ -n "$AWAITING_TASK" ]; then
@@ -130,24 +124,44 @@ if [ "$JSON" -eq 1 ]; then
   exit 0
 fi
 
-# One line per bucket, each naming its own next action, because they are
-# different actions. Counts are printed even when they are 1: "1 run" and
-# "13 runs" call for different reactions and the old line could not tell them
-# apart.
+# One line per bucket, each naming its own next action. Counts are printed
+# even when they are 1, because "1 run" and "13 runs" call for different
+# reactions.
+LANG_CODE=$(ma_output_lang 2>/dev/null || echo en)
 if [ "$AWAITING_N" -gt 0 ]; then
-  printf 'multi-agent: %s run(s) waiting on you, oldest %s at Phase %s - `%s` to finish.\n' \
-    "$AWAITING_N" "$AWAITING_TASK" "$AWAITING_PHASE" "$(ma_command resume)"
+  if [ "$LANG_CODE" = "tr" ]; then
+    printf 'multi-agent: %s çalışma seni bekliyor, en eskisi %s (Phase %s). Bitirmek için `%s`.\n' \
+      "$AWAITING_N" "$AWAITING_TASK" "$AWAITING_PHASE" "$(ma_command resume)"
+  else
+    printf 'multi-agent: %s run(s) waiting on you, oldest %s at Phase %s - `%s` to finish.\n' \
+      "$AWAITING_N" "$AWAITING_TASK" "$AWAITING_PHASE" "$(ma_command resume)"
+  fi
 fi
 if [ "$DIED_N" -gt 0 ]; then
-  printf 'multi-agent: %s run(s) stopped mid-development, oldest %s at Phase %s - `%s` or `%s`.\n' \
-    "$DIED_N" "$DIED_TASK" "$DIED_PHASE" "$(ma_command resume)" "$(ma_command kill)"
+  if [ "$LANG_CODE" = "tr" ]; then
+    printf 'multi-agent: %s çalışma geliştirme sırasında durdu, en eskisi %s (Phase %s). Devam için `%s`, silmek için `%s`.\n' \
+      "$DIED_N" "$DIED_TASK" "$DIED_PHASE" "$(ma_command resume)" "$(ma_command kill)"
+  else
+    printf 'multi-agent: %s run(s) stopped mid-development, oldest %s at Phase %s - `%s` or `%s`.\n' \
+      "$DIED_N" "$DIED_TASK" "$DIED_PHASE" "$(ma_command resume)" "$(ma_command kill)"
+  fi
 fi
 if [ "$UNSTARTED_N" -gt 0 ]; then
-  printf 'multi-agent: %s run(s) left at a Phase 0 question, nothing built - `%s --abandoned` to clear.\n' \
-    "$UNSTARTED_N" "$(ma_command garbage-collect)"
+  if [ "$LANG_CODE" = "tr" ]; then
+    printf 'multi-agent: %s çalışma Phase 0 sorusunda kaldı, hiçbir şey üretilmedi. Temizlemek için `%s --abandoned`.\n' \
+      "$UNSTARTED_N" "$(ma_command garbage-collect)"
+  else
+    printf 'multi-agent: %s run(s) left at a Phase 0 question, nothing built - `%s --abandoned` to clear.\n' \
+      "$UNSTARTED_N" "$(ma_command garbage-collect)"
+  fi
 fi
 if [ "$REVIEW_DUE" -eq 1 ]; then
-  printf 'multi-agent: %s open pipeline observation(s), last review %s+ days ago - `%s backlog` when convenient.\n' \
-    "$OPEN_OBS" "$REVIEW_DAYS" "$(ma_command refactor)"
+  if [ "$LANG_CODE" = "tr" ]; then
+    printf 'multi-agent: %s açık pipeline gözlemi var, son inceleme %s+ gün önce. Uygun olduğunda `%s backlog`.\n' \
+      "$OPEN_OBS" "$REVIEW_DAYS" "$(ma_command refactor)"
+  else
+    printf 'multi-agent: %s open pipeline observation(s), last review %s+ days ago - `%s backlog` when convenient.\n' \
+      "$OPEN_OBS" "$REVIEW_DAYS" "$(ma_command refactor)"
+  fi
 fi
 exit 0

package/pipeline/scripts/check-derived-drift.mjs

@@ -5,7 +5,7 @@
 // This exists because the resolution order is the whole correctness question,
 // and prose in a skill file cannot be tested. The step used to read the
 // installed plugin cache and call the answer authoritative. A cache is only as
-// fresh as the last `claude marketplace update`, so it reported "up to date"
+// fresh as the last `claude plugin marketplace update`, so it reported "up to date"
 // for a derivation that was four releases behind, and the version it had been
 // recorded from was the wrong manifest besides.
 //

package/pipeline/scripts/doctor.mjs

@@ -476,6 +476,27 @@ function checkHookCoverage() {
     );
     return;
   }
+  const seen = new Set();
+  const twice = new Set();
+  for (const [event, groups] of Object.entries(have?.hooks || {})) {
+    for (const group of Array.isArray(groups) ? groups : []) {
+      for (const h of group?.hooks || []) {
+        if (typeof h?.command !== "string") continue;
+        const key = `${event} ${group.matcher || ""} ${h.command}`;
+        if (seen.has(key)) twice.add(`${h.command.split("/").pop()} on ${group.matcher || event}`);
+        seen.add(key);
+      }
+    }
+  }
+  if (twice.size) {
+    report(
+      "hook-coverage",
+      "WARN",
+      `registered twice, so it runs twice per call: ${[...twice].join(", ")}`,
+      `run ${hostCommand("update", HERE)}`,
+    );
+    return;
+  }
   ok("hook-coverage");
 }
 

package/pipeline/scripts/plugin-refresh.sh

@@ -0,0 +1,92 @@
+#!/usr/bin/env bash
+#
+# plugin-refresh.sh  -  bring the stack plugins to the latest published version.
+#
+# Two steps, because they are two different things: `claude plugin marketplace
+# update` refreshes the catalog, and each installed plugin then needs its own
+# `claude plugin update` in the scope it was installed in. A project or local
+# install is updated from inside its own project, because that is how the CLI
+# tells two projects' installs apart. The catalog refresh alone leaves every
+# installed plugin where it was.
+#
+# Usage: plugin-refresh.sh <owner> [--add-only]
+#   <owner>      GitHub owner of the multi-agent-plugins repository
+#   --add-only   make sure the marketplace is configured, update nothing
+#
+# The CLI's own error output is passed through, never discarded.
+# Exit codes: 0 = done (or no claude CLI on this machine), 1 = a step failed
+
+set -uo pipefail
+
+OWNER="${1:-}"
+MODE="${2:-}"
+MP="multi-agent-plugins"
+CLAUDE="${CLAUDE_BIN:-claude}"
+
+if [ -z "$OWNER" ]; then
+  echo "usage: plugin-refresh.sh <owner> [--add-only]" >&2
+  exit 1
+fi
+
+if ! command -v "$CLAUDE" >/dev/null 2>&1; then
+  echo "  -> claude CLI not found; stack plugins not refreshed"
+  exit 0
+fi
+
+FAILED=0
+
+if ! "$CLAUDE" plugin marketplace list 2>/dev/null | grep -q "$MP"; then
+  if "$CLAUDE" plugin marketplace add "$OWNER/$MP"; then
+    echo "  -> marketplace $MP added"
+  else
+    echo "  -> could not add marketplace $OWNER/$MP" >&2
+    exit 1
+  fi
+fi
+
+[ "$MODE" = "--add-only" ] && exit 0
+
+if "$CLAUDE" plugin marketplace update "$MP"; then
+  echo "  -> marketplace $MP refreshed"
+else
+  echo "  -> marketplace $MP refresh failed" >&2
+  FAILED=1
+fi
+
+INSTALLED=$("$CLAUDE" plugin list --json 2>/dev/null | node -e '
+let s = "";
+process.stdin.on("data", (d) => (s += d)).on("end", () => {
+  let rows = [];
+  try { rows = JSON.parse(s); } catch { process.exit(0); }
+  const seen = new Set();
+  for (const r of Array.isArray(rows) ? rows : []) {
+    if (typeof r.id !== "string" || !r.id.endsWith("@" + process.argv[1])) continue;
+    const scope = r.scope || "user";
+    const dir = scope === "project" || scope === "local" ? r.projectPath || "" : "";
+    if ((scope === "project" || scope === "local") && !dir) continue;
+    const key = [r.id, scope, dir].join("\t");
+    if (seen.has(key)) continue;
+    seen.add(key);
+    console.log(key);
+  }
+});' "$MP")
+
+while IFS=$'\t' read -r ID SCOPE DIR; do
+  [ -n "$ID" ] || continue
+  WHERE="$SCOPE${DIR:+ $DIR}"
+  if [ -n "$DIR" ] && [ ! -d "$DIR" ]; then
+    echo "  -> $ID ($WHERE) skipped, project directory is gone"
+    continue
+  fi
+  if (cd "${DIR:-$PWD}" && "$CLAUDE" plugin update "$ID" --scope "$SCOPE"); then
+    echo "  -> $ID ($WHERE) updated"
+  else
+    echo "  -> $ID ($WHERE) update failed" >&2
+    FAILED=1
+  fi
+done <<EOF
+$INSTALLED
+EOF
+
+[ "$FAILED" -eq 0 ] || exit 1
+exit 0

package/pipeline/scripts/usage-register.mjs

@@ -49,10 +49,12 @@ import { join } from "node:path";
 import { homedir, hostname } from "node:os";
 import { pathToFileURL } from "node:url";
 import { hostCommand } from "../lib/user-facing.mjs";
+import { canonicalEndpoint } from "../lib/usage-endpoint.mjs";
 import { ghTokenFor, resolveReportingLogin } from "./usage-identity.mjs";
 import { lockTimeoutFrom, updateJsonFileSync } from "../lib/json-file-lock.mjs";
 
 const DEFAULT_ENDPOINT = "https://mmerterden.com/api/usage/ingest";
+const BUILT_IN_HOSTS = ["mmerterden.com"];
 const TIMEOUT_MS = 10000;
 
 function readJson(file) {
@@ -118,7 +120,7 @@ export function decideRegistration(
 
 /** The /register URL derived from whatever endpoint the prefs name. */
 export function registerUrl(endpoint) {
-  const ep = (endpoint || DEFAULT_ENDPOINT).trim().replace(/\/+$/, "");
+  const ep = canonicalEndpoint((endpoint || DEFAULT_ENDPOINT).trim()).replace(/\/+$/, "");
   return ep.endsWith("/ingest") ? `${ep.slice(0, -"/ingest".length)}/register` : `${ep}/register`;
 }
 
@@ -132,7 +134,7 @@ function storedToken(prefs, credStore) {
 
 /**
  * May the gh token be sent to `url`? The token is a full GitHub credential, so
- * it only goes over https to the default reporting host or a host the prefs
+ * it only goes over https to the reporting site's own hosts or a host the prefs
  * name in `usageLog.trustedHosts`; http only to loopback (local development).
  */
 export function registerUrlAllowed(url, trustedHosts = []) {
@@ -144,9 +146,7 @@ export function registerUrlAllowed(url, trustedHosts = []) {
   }
   if (u.protocol === "http:") return ["127.0.0.1", "localhost", "[::1]"].includes(u.hostname);
   if (u.protocol !== "https:") return false;
-  const trusted = new Set(
-    [new URL(DEFAULT_ENDPOINT).hostname, ...trustedHosts].map((h) => String(h).toLowerCase()),
-  );
+  const trusted = new Set([...BUILT_IN_HOSTS, ...trustedHosts].map((h) => String(h).toLowerCase()));
   return trusted.has(u.hostname.toLowerCase());
 }
 
@@ -355,7 +355,7 @@ async function main(argv) {
 
   const optedOut = prefs.global?.usageLog?.optOut === true;
   const entry = `${process.env.USER || user}_Usage_Ingest_Token`;
-  const set = spawnSync("bash", [credStore, "set", entry], {
+  const set = spawnSync("bash", [credStore, "set", entry, "-"], {
     input: token,
     encoding: "utf8",
     timeout: 10000,

package/pipeline/scripts/usage-report.mjs

@@ -28,6 +28,7 @@
  */
 
 import { readFileSync, existsSync, readdirSync, appendFileSync } from "fs";
+import { canonicalEndpoint } from "../lib/usage-endpoint.mjs";
 import { homedir, platform } from "os";
 import { dirname, join } from "path";
 import { fileURLToPath, pathToFileURL } from "url";
@@ -339,10 +340,10 @@ function commandOf(state, mode) {
   return mode;
 }
 
-// A repository name is the user's, not ours. It put a corporate project into
-// the maintainer's database, which is the disclosure this telemetry must not
-// make - and the panel never needed the name, only the ability to tell one
-// project's runs from another's. A digest does that and cannot be read back.
+// A repository name is the user's, not ours, and the panel never needs it:
+// only the ability to tell one project's runs from another's. A digest keeps
+// the name out of the record. It is not secret: anyone who can guess a name
+// can hash it and compare.
 function projectDigest(name) {
   if (typeof name !== "string" || !name.trim()) return null;
   return `p${createHash("sha256").update(name.trim()).digest("hex").slice(0, 8)}`;
@@ -721,7 +722,7 @@ async function main() {
   const args = parseArgs(process.argv.slice(2));
   if (process.env.MULTI_AGENT_SMOKE && !args.dryRun) return;
   const prefs = resolvePrefs();
-  const endpoint = prefs.endpoint || ENDPOINT_DEFAULT;
+  const endpoint = canonicalEndpoint(prefs.endpoint || ENDPOINT_DEFAULT);
 
   // Gate BEFORE touching the keychain: optOut is a hard block, and reading the
   // secret for a disabled feature is both wasteful and can raise a macOS

package/pipeline/skills/.skill-manifest.json

@@ -1,6 +1,6 @@
 {
   "schemaVersion": "1.0.0",
-  "generatedAt": "2026-09-25T08:57:50Z",
+  "generatedAt": "2026-09-25T15:14:42Z",
   "skillCount": 218,
   "entries": [
     {
@@ -45,7 +45,7 @@
     },
     {
       "path": "shared/core/multi-agent-channels/SKILL.md",
-      "sha256": "e52daada5aba4f772018d4cab508acc87b9b1c2686e1f5a770c4843f76c74208"
+      "sha256": "03f7c648e1938f3d6d7c48d4c1c01f2bd734cdce873ec5203120b5be3ae761db"
     },
     {
       "path": "shared/core/multi-agent-complaint-analysis/SKILL.md",
@@ -97,7 +97,7 @@
     },
     {
       "path": "shared/core/multi-agent-jira/SKILL.md",
-      "sha256": "c3e75df9f353bfeb0eb33632fab409721f03aab9c901102502d328eea93a6092"
+      "sha256": "d7e23fd1565b6640d33c23cce4d025660212d9d3f3ab80a43369ad80ae8be080"
     },
     {
       "path": "shared/core/multi-agent-kill/SKILL.md",
@@ -133,7 +133,7 @@
     },
     {
       "path": "shared/core/multi-agent-refactor/SKILL.md",
-      "sha256": "dedcdb18c961b0ddf87d1f3308e9aba94cc31534cc4263503098fbdd5dd292e1"
+      "sha256": "889c17f800b83d5c5d13d097eff63f28dc8d7869679bfcdb72efecd8b1d66391"
     },
     {
       "path": "shared/core/multi-agent-research/SKILL.md",
@@ -205,7 +205,7 @@
     },
     {
       "path": "shared/core/multi-agent-stack/SKILL.md",
-      "sha256": "827d1582ee536ed65660574f24687b17b8b0a2a76f9b69ab03c9bb184a7712bb"
+      "sha256": "ac964cd00dd6ed06512530489d867efbc51d7d7614265f9aa022679325d5defe"
     },
     {
       "path": "shared/core/multi-agent-status/SKILL.md",
@@ -253,11 +253,11 @@
     },
     {
       "path": "shared/core/multi-agent-update/SKILL.md",
-      "sha256": "b58574eb82a28c1657d30524b528a55a8d1f3e2eadd7b23f85525b2aca243c4b"
+      "sha256": "d96e18b70bab0a601057f75f212dd4fb563db0cb250607c770ce1c7a1df4b1d0"
     },
     {
       "path": "shared/core/multi-agent/SKILL.md",
-      "sha256": "b738b5055f4eb9a37934730ceddec8f02058c9a5e5c3d99158de1b4634e0d2a6"
+      "sha256": "3bfaa4f0743aa1ab8d7f2bec4edb6b3c759f6c7a890535b2f0e1606f2ce7547d"
     },
     {
       "path": "shared/external/accessibility-compliance-accessibility-audit/SKILL.md",