feat(gitlab): gl-mrs MR triage board + watch-mine supervisor (#275)

* feat(gitlab): gl-mrs MR triage board + watch-mine supervisor

gl-mrs lists MRs (default author=@me), sorted failing-first, each enriched
in parallel with what you'd otherwise round-trip for: pipeline status
(a failure shows the failed job name = the failure class), approval state,
age, diff size, watch-state cross-reference, and conflict/draft/threads
flags. Actionable footer points at the first failing-and-unwatched MR.

"Mine" is just the default filter, not a special op — enumeration is a
platform concern, kept out of the generic watch preset.

watch-mine.sh: glue between a list-mine op and the watch preset — runs the
feed op, extracts ids, spawns one watcher each. Idempotent, safe on /loop.
gl-mrs:failed,iids | watch-mine turns "ping me when any of my MRs goes red"
into one looped command.

Enrich knobs (enrich_workers/enrich_cap/per_page) passed through as
SUPERTOOL_* env. 44 unit tests; full suite 3135 passed.

Refs #274 (gh-prs twin)

Co-Authored-By: Max <noreply>

* fix(gl-mrs): review fixes — iids None-guard, fd leak, _age skew, main() tests

Addresses PR #275 review:
- iids output guards against a missing iid (no stray "None" into the
  watch-mine feed)
- _watched_iids closes the PID file via context manager (no fd leak)
- _age guards future timestamps (clock skew) → "now" instead of "-1d";
  datetime import hoisted to module top
- +9 tests: _age skew/bad-iso, _flags True-suppression, and main()
  integration (iids output, missing-iid skip, failed filter, glab error,
  bad JSON) — covers the watch-mine.sh contract

53 gl-mrs tests pass.

Co-Authored-By: Max <noreply>

Author: fdaviddpt

.supertool.example.json

@@ -274,6 +274,15 @@
       "lines": 80,
       "error_patterns": "ERROR,FAIL,Fatal"
     },
+    "gl-mrs": {
+      "cmd": "python3 {path}gitlab/mrs.py {args}",
+      "timeout": 60,
+      "description": "MR triage board. Override the gitlab preset's defaults to tune pipeline enrichment — extra keys pass through as SUPERTOOL_ENRICH_WORKERS / SUPERTOOL_ENRICH_CAP / SUPERTOOL_PER_PAGE.",
+      "example": "gl-mrs:author=@me,failed",
+      "enrich_workers": 8,
+      "enrich_cap": 40,
+      "per_page": 50
+    },
     "_hidden_internal": {
       "cmd": "echo internal-helper",
       "description": "Hidden from `./supertool ops` output via status:0 — useful for ops that exist for aliases or scripts but shouldn't clutter the public listing.",

CHANGELOG.md

@@ -9,6 +9,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Added
 
+- **`gl-mrs` — MR triage board.** Lists MRs (default `author=@me`), sorted failing-first then stalest, each enriched in parallel with everything you'd otherwise round-trip for: pipeline status (a failure shows the failed **job name** — `phpstan2`, `test_unit_dpt` — i.e. the failure class, pre-empting a `gl-job` call), approval state, age, diff size, watch-state cross-reference (which MRs already have a live `watch` poller), and `conflict`/`draft`/`threads` flags. Footer points at the first failing-and-unwatched MR. Filters: `author`/`reviewer`/`assignee`/`label`/`milestone`/`state`/`per`. Flags: `nopipe`, `iids` (bare id list), `failed` (only failing). Tunable via `enrich_workers`/`enrich_cap`/`per_page`. "Mine" is just the default filter, not a special op — enumeration is a platform concern, kept out of the generic `watch` preset. See [docs/presets/gitlab.md](docs/presets/gitlab.md).
+- **`presets/watch/watch-mine.sh` — watch-a-whole-query supervisor.** Glue between a "list mine" op (`gl-mrs`/`gh-prs`) and the `watch` preset: runs the feed op, extracts ids, spawns one watcher each. Idempotent, so it's safe on `/loop`. `gl-mrs:failed,iids | watch` turns "ping me when any of my MRs goes red" into one looped command. See [docs/presets/watch.md](docs/presets/watch.md).
 - **`watch` preset — background event pollers with async wake into Claude Code.** New ops `watch:SOURCE:ID[:only=...]`, `unwatch:SOURCE:ID`, `watches`. Each invocation forks a detached poller that emits state-change events to a UDS socket, a status file, and macOS Notification Center. Source-plugin contract makes new sources ~50 lines. Bundled sources: `github-pr` (PR state, checks, reviews, comments, merge/close, conflicts) and `gitlab-mr` (MR state, pipeline, merge/close, conflicts). Closes [#165](https://github.com/Digital-Process-Tools/claude-supertool/issues/165). See [docs/presets/watch.md](docs/presets/watch.md).
 - **`notifiers/claude-channel/` — MCP channel server for async wake.** TypeScript / Bun server that binds the watch UDS socket and pushes events into a running Claude Code session via the Channels feature (research preview, requires Claude Code v2.1.80+). Events arrive as `<channel source="claude-channel" watcher_source="<source>" id="<id>" event="<event>" ...>` tags. UDS file mode 0600 + localhost-only auth model. Launch with `claude --dangerously-load-development-channels server:claude-channel`. See [notifiers/claude-channel/README.md](notifiers/claude-channel/README.md).
 - **`glob` brace expansion.** `glob:src/**/*.{json,xml}` now fans out into `*.json` + `*.xml` (shell/fd/ripgrep semantics) and dedupes results, instead of silently returning 0 files. Supports multiple groups (`{a,b}.{x,y}` → 4) and nesting (`{a,b{1,2}}` → 3). Patterns without braces are unchanged. Closes [#161](https://github.com/Digital-Process-Tools/claude-supertool/issues/161).

README.md

@@ -32,6 +32,7 @@ supertool 'read:src/Module.py' 'read:src/Auth.py' 'grep:TODO:src/:20' 'map:src/'
 
 - **`git-status`** — branch + tracking + ahead/behind + dirty files + open MR/PR + suggested next step. One call, decision ready.
 - **`gl-mr:NUMBER`** / **`gh-pr:NUMBER`** — full MR/PR dashboard: branch, pipeline, reviewer, approval, diff stat, comments. Replaces 4-5 `glab`/`gh` calls.
+- **`gl-mrs`** — MR triage board: your open MRs + per-MR pipeline status + which already have a `watch` poller running + an actionable footer. Pairs with `watch` to auto-watch every failing MR.
 - **`claude-log-summary:UUID`** — model, duration, tool calls, tokens, cache hit %, errors-by-tool. Audit your own runs.
 
 That's a sample. supertool ships ~40 ops out of the box (built-ins + `gitlab` / `github` / `git` / `claude-log` presets) — add your own and you're past 60 fast.

docs/presets/gitlab.md

@@ -12,6 +12,7 @@ GitLab ops via the `glab` CLI. Replaces the 3-5 separate `glab` calls needed to
 |----|--------|-----------------|
 | `gl-issue` | `gl-issue:NUMBER[:full]` | Issue metadata, description, comments (truncated by default), related MRs. `:full` disables truncation |
 | `gl-mr` | `gl-mr:NUMBER_OR_BRANCH[:status]` | MR dashboard: branch, pipeline, reviewer/approval state, linked issue, diff stat, comments. `:status` returns slim merge-state only |
+| `gl-mrs` | `gl-mrs[:filters,flags]` | MR triage board, sorted failing-first then stalest. Per MR (enriched in parallel): pipeline status (a failure shows the failed **job name** = the failure class), approval state, age, diff size, watch-state cross-reference, and `conflict`/`draft`/`threads` flags — plus an actionable footer. Filters (comma-sep): `author`/`reviewer`/`assignee`/`label`/`milestone`/`state`/`per`. Flags: `nopipe` (skip enrichment), `iids` (bare id list), `failed` (only failing) |
 | `gl-pipeline` | `gl-pipeline:NUMBER` | Pipeline job list grouped by stage with pass/fail status and failed job IDs |
 | `gl-job` | `gl-job:NUMBER[:raw[:START[:END]]]` | Job failure detail: MR context + error pattern search + log tail. `:raw` dumps the full trace; `:raw:START:END` slices lines (1-indexed, inclusive) |
 
@@ -38,8 +39,33 @@ Returns approval state, pipeline status, diff stat, and all comments in one call
 ```
 Gets the full issue context and checks whether an MR already exists for the branch.
 
+**Triage your open MRs:**
+```bash
+./supertool 'gl-mrs'                          # my MRs: pipeline, approval, age, watch-state
+./supertool 'gl-mrs:reviewer=@me'             # MRs I'm reviewing
+./supertool 'gl-mrs:author=@me,failed,iids'   # bare ids of my failing MRs
+```
+The last form feeds the [`watch`](watch.md) supervisor: pipe failing-MR ids straight into background pollers.
+
 ## Configuration
 
+`gl-mrs` enrichment is tunable (parallelism, how many MRs to enrich, page size):
+
+```json
+{
+  "ops": {
+    "gl-mrs": {
+      "cmd": "python3 {path}gitlab/mrs.py {args}",
+      "enrich_workers": 8,
+      "enrich_cap": 40,
+      "per_page": 50
+    }
+  }
+}
+```
+
+The list endpoint omits pipeline, approval, and diff data, so each MR costs a couple of `glab api` calls (MR detail + approvals, plus failed-job names for failing MRs). `enrich_workers` threads those calls; `enrich_cap` bounds how many MRs get enriched; `per_page` bounds how many are listed.
+
 `gl-job` error pattern search and log tail length are configurable:
 
 ```json

docs/presets/watch.md

@@ -26,6 +26,23 @@ Composable in one batched call:
 ./supertool 'watch:github-pr:179' 'watch:github-pr:180' 'watches'
 ```
 
+## Watch many from a query — `watch-mine.sh`
+
+`watch:SOURCE:ID` watches one id. To watch *every id a query returns* — e.g. all your failing MRs — pair it with a list op via the bundled supervisor `presets/watch/watch-mine.sh`. It runs a "list mine" op (`gl-mrs`/`gh-prs`), extracts the ids, and spawns one watcher each. Idempotent (the `watch` op skips ids already watched), so it's safe on a loop:
+
+```bash
+# default: my failing GitLab MRs → gitlab-mr watchers
+bash presets/watch/watch-mine.sh
+
+# re-sync every 5 min from inside Claude Code
+/loop 5m bash presets/watch/watch-mine.sh
+
+# any feed op + source — e.g. my failing GitHub PRs (once a gh-prs op exists)
+bash presets/watch/watch-mine.sh 'gh-prs:author=@me,failed,iids' github-pr
+```
+
+Args: `$1` feed op (default `gl-mrs:author=@me,failed,iids`), `$2` watch source (default `gitlab-mr`), `$3` notify events (default `pipeline_failed,merged`). The separation is deliberate — the list op owns *what's mine* (a platform concern), the watch preset stays generic. The feed op just has to emit bare ids (the `iids` flow); only `gl-mrs` ships today, a `gh-prs` twin is the obvious next one.
+
 ## Bundled sources
 
 | Source | Polls | Events |

presets/gitlab.json

@@ -14,6 +14,15 @@
       "description": "MR review: branch, pipeline, approval, linked issue, diff stat, comments. :status = slim merge-state only",
       "syntax": "gl-mr:NUMBER_OR_BRANCH[:status]"
     },
+    "gl-mrs": {
+      "cmd": "{python} {path}gitlab/mrs.py {args}",
+      "timeout": 60,
+      "description": "MR triage board: list MRs (default author=@me), sorted failing-first. Per MR: pipeline status (failed→job name), approval state, age, diff size, watch-state, conflict/draft/threads flags + actionable footer. Filters (comma-sep): author/reviewer/assignee/label/milestone/state. Flags: nopipe, iids, failed. Enrichment calls run in parallel.",
+      "syntax": "gl-mrs[:author=@me,reviewer=@me,state=opened,failed,nopipe,iids]",
+      "enrich_workers": 8,
+      "enrich_cap": 40,
+      "per_page": 50
+    },
     "gl-issue-create": {
       "cmd": "{python} {path}gitlab/issue_create.py {arg}",
       "timeout": 30,