fix(office-hours): session writer was writing to legacy file

CHANGELOG.md

@@ -1,5 +1,34 @@
 # Changelog
 
+## [1.43.4.0] - 2026-05-24
+
+## **Returning `/office-hours` users now skip the introduction-tier closing pitch on their second visit.**
+## **The v1.0.0.0 writer/reader split is closed: sessions now flow through `gstack-developer-profile --log-session`, which writes to the same file the reader reads.**
+
+The bug: `office-hours/SKILL.md.tmpl` was writing session entries to `~/.gstack/builder-profile.jsonl` (the legacy path), while the v1.0.0.0 migration had moved the read path to `~/.gstack/developer-profile.json`. The reader saw `sessions: []` every time, reported `SESSION_COUNT: 0` and `TIER: introduction`, and the welcome_back tier (designed specifically to skip the closing pitch for returning users) was unreachable. Every fresh-`$HOME` user has been stuck on this since 2026-04-18.
+
+The fix: a new `--log-session` subcommand on `bin/gstack-developer-profile` that read-modify-writes the existing `sessions[]` array. Naming matches the `gstack-*-log` family verb. Atomic mktemp+mv write. Validates input (silent skip on invalid JSON or missing required fields, matching `gstack-timeline-log:22-26`). Aggregates `signals_accumulated`, `resources_shown`, and `topics` inline. Calls `gstack-brain-enqueue` after write, mirroring `gstack-timeline-log:40`.
+
+A latent bug surfaced and also fixed: `do_read` now filters `mode:"resources"` entries when picking `LAST_PROJECT` / `LAST_ASSIGNMENT` / `LAST_DESIGN_TITLE` / `CROSS_PROJECT`. The Phase 6 resources auto-append runs after the real session in the same `/office-hours` invocation; without the filter, the resources entry would have clobbered real-session state for the user's next visit. This bug was masked by the broken writer (no writes were landing) and activated by this fix.
+
+### Itemized changes
+
+### Added
+- `bin/gstack-developer-profile --log-session '<json>'` — appends a session entry to `developer-profile.json`'s `sessions[]` array, with validation and aggregate updates. Joins the `gstack-*-log` family naming convention.
+- `test/static-no-legacy-writes.test.ts` — static-grep invariant walking every skill dir to prevent future writers from regressing onto the legacy file.
+- `test/gstack-developer-profile.test.ts` — +8 tests covering the regression (read-write-read promotes to welcome_back), aggregation (signals, resources_shown, topics dedup), validation (silent skip on invalid input), `ts` injection/preservation, and the `do_read` mode filter.
+
+### Fixed
+- `/office-hours` returning users no longer get pitched as first-timers on every visit. The welcome_back tier is now reachable for the first time since v1.0.0.0.
+- `do_read` no longer picks `LAST_PROJECT` / `LAST_ASSIGNMENT` / `LAST_DESIGN_TITLE` from a Phase 6 resources auto-append entry.
+- `office-hours/SKILL.md.tmpl` writer call sites (lines 490 + 893) now use the new subcommand. Regenerated `office-hours/SKILL.md`.
+
+### Affected users
+Stranded entries in `~/.gstack/builder-profile.jsonl` from the broken-writer period are not recovered automatically by this PR, by design. On the next `/office-hours` run, the first new session counts toward the welcome_back tier; past data stays in the legacy file (still readable by other tools during deprecation). Most pre-existing users have only a handful of stranded sessions.
+
+### For contributors
+See `docs/designs/FIX_1671_PROFILE_MIGRATION.md` for the full design including what's intentionally NOT in this fix (no reconcile path, no schema bump, no mkdir-locks — each justified against codebase patterns). Related but separate PRs: #1671 RC2 (autoplan timeline rollup), #1671 RC3 (cross-identity scope opt-in). Unblocks issue #1139's proposed fix #3 (suppress repeated YC apply suggestion via prior-session detection).
+
 ## [1.43.3.0] - 2026-05-21
 
 ## **Headed Chromium embedded by external supervisors stops auto-shutting-down after 30 minutes of HTTP idle.**

VERSION

@@ -1 +1 @@
-1.43.3.0
+1.43.4.0

bin/gstack-developer-profile

@@ -17,6 +17,9 @@
 #   --check-mismatch    detect meaningful gaps between declared and observed.
 #   --migrate           migrate builder-profile.jsonl → developer-profile.json.
 #                       Idempotent; archives the source file on success.
+#   --log-session    append a session entry (from /office-hours) to
+#                       sessions[] and update aggregates. Required fields:
+#                       date, mode. Silent skip on invalid input.
 #
 # Profile file: ~/.gstack/developer-profile.json (unified schema — see
 # docs/designs/PLAN_TUNING_V0.md). Event file: ~/.gstack/projects/{SLUG}/
@@ -154,6 +157,65 @@ ensure_profile() {
 EOF
 }
 
+# -----------------------------------------------------------------------
+# Record session: append a session entry from /office-hours to sessions[]
+# and update aggregates (signals_accumulated, resources_shown, topics).
+# Fix for #1671: the writer side of the v1.0.0.0 migration. Reader and
+# writer now share the same file.
+# Silent skip on invalid input (matches gstack-timeline-log:22-26 pattern).
+# -----------------------------------------------------------------------
+do_log_session() {
+  local INPUT="${1:-}"
+  if [ -z "$INPUT" ]; then
+    return 0
+  fi
+
+  # Validate: input must be parseable JSON with required fields (date, mode).
+  if ! printf '%s' "$INPUT" | bun -e "
+    const j = JSON.parse(await Bun.stdin.text());
+    if (!j.date || !j.mode) process.exit(1);
+  " 2>/dev/null; then
+    return 0
+  fi
+
+  ensure_profile
+
+  local TMPOUT
+  TMPOUT=$(mktemp "$GSTACK_HOME/developer-profile.json.XXXXXX.tmp")
+  trap 'rm -f "$TMPOUT"' EXIT
+
+  PROFILE_FILE_PATH="$PROFILE_FILE" RECORD_INPUT="$INPUT" TMPOUT_PATH="$TMPOUT" bun -e "
+    const fs = require('fs');
+    const entry = JSON.parse(process.env.RECORD_INPUT);
+    if (!entry.ts) entry.ts = new Date().toISOString();
+
+    const profile = JSON.parse(fs.readFileSync(process.env.PROFILE_FILE_PATH, 'utf-8'));
+    profile.sessions = profile.sessions || [];
+    profile.sessions.push(entry);
+
+    profile.signals_accumulated = profile.signals_accumulated || {};
+    for (const s of (entry.signals || [])) {
+      profile.signals_accumulated[s] = (profile.signals_accumulated[s] || 0) + 1;
+    }
+
+    profile.resources_shown = profile.resources_shown || [];
+    const resSet = new Set(profile.resources_shown);
+    for (const r of (entry.resources_shown || [])) resSet.add(r);
+    profile.resources_shown = Array.from(resSet);
+
+    profile.topics = profile.topics || [];
+    const topicSet = new Set(profile.topics);
+    for (const t of (entry.topics || [])) topicSet.add(t);
+    profile.topics = Array.from(topicSet);
+
+    fs.writeFileSync(process.env.TMPOUT_PATH, JSON.stringify(profile, null, 2));
+  "
+
+  mv "$TMPOUT" "$PROFILE_FILE"
+  trap - EXIT
+  "$SCRIPT_DIR/gstack-brain-enqueue" "developer-profile.json" 2>/dev/null &
+}
+
 # -----------------------------------------------------------------------
 # Read: emit legacy KEY: VALUE output for /office-hours compat.
 # -----------------------------------------------------------------------
@@ -168,14 +230,19 @@ do_read() {
     else if (count >= 4) tier = 'regular';
     else if (count >= 1) tier = 'welcome_back';
 
-    const last = sessions[count - 1] || {};
-    const prev = sessions[count - 2] || {};
+    // LAST_* / CROSS_PROJECT must reflect real sessions, not resource-tracking
+    // events (the Phase 6 auto-append). Without this filter, a session's
+    // resources entry written immediately after the real session would clobber
+    // LAST_PROJECT/LAST_ASSIGNMENT/LAST_DESIGN_TITLE.
+    const realSessions = sessions.filter(e => e.mode !== 'resources');
+    const last = realSessions[realSessions.length - 1] || {};
+    const prev = realSessions[realSessions.length - 2] || {};
     const crossProject = prev.project_slug && last.project_slug
       ? prev.project_slug !== last.project_slug
       : false;
 
-    const designs = sessions.map(e => e.design_doc || '').filter(Boolean);
-    const designTitles = sessions
+    const designs = realSessions.map(e => e.design_doc || '').filter(Boolean);
+    const designTitles = realSessions
       .map(e => (e.design_doc ? (e.project_slug || 'unknown') : ''))
       .filter(Boolean);
 
@@ -441,6 +508,7 @@ case "$CMD" in
   --vibe) do_vibe ;;
   --check-mismatch) do_check_mismatch ;;
   --migrate) do_migrate ;;
+  --log-session) do_log_session "$@" ;;
   --help|-h) sed -n '1,/^set -euo/p' "$0" | sed 's|^# \?||' ;;
   *)
     echo "gstack-developer-profile: unknown subcommand '$CMD'" >&2

docs/designs/FIX_1671_PROFILE_MIGRATION.md

@@ -0,0 +1,81 @@
+# Fix #1671: `/office-hours` always reports SESSION_COUNT: 0
+
+**Status:** SHIPPED
+**Branch:** fix-1671-profile-migration
+**Date:** 2026-05-23
+**Issue:** https://github.com/garrytan/gstack/issues/1671
+**Original PR that introduced the bug:** garrytan/gstack#1039 / commit `0a803f9` / v1.0.0.0 / 2026-04-18
+
+## The problem
+
+`/office-hours` reports `SESSION_COUNT: 0` and `TIER: introduction` on every invocation, even for users who have run the skill many times. The `welcome_back` tier (`bin/gstack-developer-profile:165-169`) that exists to skip the closing pitch for returning users is unreachable. Live ~5 weeks on every fresh-`$HOME` user since v1.0.0.0.
+
+## Root cause
+
+The v1.0.0.0 migration moved the read path to `~/.gstack/developer-profile.json` but left the writer in `office-hours/SKILL.md.tmpl` writing to the legacy `~/.gstack/builder-profile.jsonl`. The `ensure_profile` stub created on first read has `sessions: []`; subsequent writes go to a file the reader never re-reads. Reader and writer disagree on storage.
+
+Full root-cause analysis (including RC2/RC3 follow-ups): https://github.com/garrytan/gstack/issues/1671
+
+## The fix
+
+Make the writer use the same file the reader does.
+
+### Changes
+
+1. **`bin/gstack-developer-profile`** — add `--log-session '<json>'` subcommand:
+   - Validates required fields (`date`, `mode`), silent-skip on invalid input (matches `bin/gstack-timeline-log:22-26`).
+   - Reads existing `developer-profile.json` via `bun -e`.
+   - Appends entry to `sessions[]`. Updates `signals_accumulated` (per-signal-string increment, same as `do_migrate:67-69`), unions `resources_shown` and `topics`.
+   - Atomic mktemp+mv write (matches existing pattern at line 54).
+   - Calls `gstack-brain-enqueue "developer-profile.json"` after write, mirroring `bin/gstack-timeline-log:40`.
+
+2. **`bin/gstack-developer-profile:do_read`** — filter `mode:"resources"` entries when picking LAST_PROJECT / LAST_ASSIGNMENT / LAST_DESIGN_TITLE / CROSS_PROJECT / DESIGN_*. The Phase 6 resources auto-append happens after the real session in the same /office-hours invocation; without the filter, that resources entry clobbers real-session state for the user's next session. Latent bug that was masked by the broken writer; activated by the fix.
+
+3. **`office-hours/SKILL.md.tmpl`** — swap writers at lines 490 and 893:
+   - From: `echo '{...}' >> "$GSTACK_STATE_ROOT/builder-profile.jsonl"`
+   - To: `~/.claude/skills/gstack/bin/gstack-developer-profile --log-session '{...}' 2>/dev/null || true`
+   - Run `bun run gen:skill-docs` to regenerate `office-hours/SKILL.md`.
+
+### What's NOT in the fix (intentionally)
+
+- **No new binary.** The owner binary for `developer-profile.json` is `gstack-developer-profile`; the writer belongs there as a subcommand. `--log-session` joins the binary's existing `--migrate` / `--derive` write-side subcommand boundary, not the `gstack-*-log` event-writer family. Verb name still matches `gstack-*-log`.
+- **No mkdir-locks.** Concurrent /office-hours calls have a read-modify-write race on `developer-profile.json`. The codebase accepts the same race in `gstack-config` (r-m-w on YAML, no lock). Not introduced by this fix; out of scope.
+- **No schema bump.** Schema stays at `schema_version: 1`. The fix doesn't change the schema, just makes the writer use it.
+- **No auto-reconcile for affected users.** Existing users with stranded `builder-profile.jsonl` entries don't get their past history auto-merged into `developer-profile.json`. On their next /office-hours run, the first new session lands in `welcome_back`; past data stays in the legacy file (still readable by other tools during deprecation). Most affected users have only a handful of stranded sessions so the loss is mostly aesthetic. Dropped the one-release-only reconcile pathway as net noise — Garry's "right-sized diff" voice.
+- **No autoplan timeline rollup (RC2).** Separate concern, separate PR.
+- **No project-scope opt-in (RC3).** Separate concern, separate PR.
+- **No gbrain glob change.** The office-hours manifest still globs `~/.gstack/builder-profile.jsonl` for context; once new writes stop landing there, the snapshot goes cold. Update in a follow-up if it becomes a UX issue.
+
+### Tests (all gate-tier, free, deterministic)
+
+1. **Regression test** in `test/gstack-developer-profile.test.ts`:
+   - Fresh `$HOME`.
+   - Run /office-hours preamble: gstack-developer-profile creates empty stub.
+   - Call `--log-session` with a startup-mode JSON.
+   - Run `--read` again. Assert `SESSION_COUNT: 1`, `TIER: welcome_back`.
+   - Fails on current main (subcommand doesn't exist). Passes with fix.
+
+2. **`do_read` mode filter test:** after recording a startup session followed by a resources entry, `--read` returns LAST_PROJECT / LAST_ASSIGNMENT / LAST_DESIGN_TITLE from the real session, not from the resources entry. RESOURCES_SHOWN still aggregates correctly.
+
+3. **Validation + aggregation tests:** `--log-session` silently skips invalid JSON / missing required fields, injects `ts` if missing, preserves user-set `ts`, correctly aggregates signals/resources/topics across multiple sessions.
+
+4. **Static-grep invariant** in `test/static-no-legacy-writes.test.ts` (new): walks every skill dir, asserts no production code path writes to `builder-profile.jsonl` except allowlisted readers (`gstack-developer-profile`, `gstack-memory-ingest.ts`, `gstack-artifacts-init`, doc files). Prevents future writers from regressing onto the legacy file.
+
+### Acceptance criteria
+
+- Second `/office-hours` invocation on a fresh `$HOME` returns `TIER: welcome_back`.
+- `bun test` passes on the touched files in isolation.
+- `bun run gen:skill-docs` produces clean diff matching the `.tmpl` edits.
+
+### Rollout
+
+- One commit. PATCH version bump per CHANGELOG style guide.
+- CHANGELOG entry written by `/ship`. User-facing voice: lead with what users experience now that they didn't before (welcome_back tier kicks in on second visit).
+
+## Follow-up TODOs
+
+- Deprecate `builder-profile.jsonl` entirely (writer + shim + memory-ingest type) after one release.
+- Fix RC2 (autoplan inlines sub-skills, bypassing their timeline-log preambles).
+- Add `GSTACK_PROFILE_SCOPE` opt-in for power users with multiple agent identities (RC3).
+- /plan-tune doesn't currently call `--derive`, so `inferred`/`gap` can drift (pre-existing, unrelated to #1671).
+- `mode:"resources"` entries inflate SESSION_COUNT under the existing tier aggregator (pre-existing, unrelated to #1671 root cause).

office-hours/SKILL.md

@@ -1537,12 +1537,9 @@ Count the signals. You'll use this count in Phase 6 to determine which tier of c
 ### Builder Profile Append
 
 After counting signals, append a session entry to the builder profile. This is the single
-source of truth for all closing state (tier, resource dedup, journey tracking).
-
-```bash
-eval "$(~/.claude/skills/gstack/bin/gstack-paths)"
-mkdir -p "$GSTACK_STATE_ROOT"
-```
+source of truth for all closing state (tier, resource dedup, journey tracking). The
+`gstack-developer-profile --log-session` binary handles its own directory creation
+and writes via atomic mktemp+mv to `~/.gstack/developer-profile.json`.
 
 Append one JSON line with these fields (substitute actual values from this session):
 - `date`: current ISO 8601 timestamp
@@ -1556,12 +1553,12 @@ Append one JSON line with these fields (substitute actual values from this sessi
 - `topics`: array of 2-3 topic keywords that describe what this session was about
 
 ```bash
-eval "$(~/.claude/skills/gstack/bin/gstack-paths)"
-echo '{"date":"TIMESTAMP","mode":"MODE","project_slug":"SLUG","signal_count":N,"signals":SIGNALS_ARRAY,"design_doc":"DOC_PATH","assignment":"ASSIGNMENT_TEXT","resources_shown":[],"topics":TOPICS_ARRAY}' >> "$GSTACK_STATE_ROOT/builder-profile.jsonl"
+~/.claude/skills/gstack/bin/gstack-developer-profile --log-session '{"date":"TIMESTAMP","mode":"MODE","project_slug":"SLUG","signal_count":N,"signals":SIGNALS_ARRAY,"design_doc":"DOC_PATH","assignment":"ASSIGNMENT_TEXT","resources_shown":[],"topics":TOPICS_ARRAY}' 2>/dev/null || true
 ```
 
-This entry is append-only. The `resources_shown` field will be updated via a second append
-after resource selection in Phase 6 Beat 3.5.
+The session entry is appended to `developer-profile.json`'s `sessions[]` array. A second
+session entry with `mode: "resources"` is appended via `--log-session` after resource
+selection in Phase 6 Beat 3.5.
 
 ---
 
@@ -2018,8 +2015,8 @@ PAUL GRAHAM ESSAYS:
 1. Log the selected resource URLs to the builder profile (single source of truth).
 Append a resource-tracking entry:
 ```bash
-eval "$(~/.claude/skills/gstack/bin/gstack-paths)"
-echo '{"date":"'"$(date -u +%Y-%m-%dT%H:%M:%SZ)"'","mode":"resources","project_slug":"'"${SLUG:-unknown}"'","signal_count":0,"signals":[],"design_doc":"","assignment":"","resources_shown":["URL1","URL2","URL3"],"topics":[]}' >> "$GSTACK_STATE_ROOT/builder-profile.jsonl"
+eval "$(~/.claude/skills/gstack/bin/gstack-slug 2>/dev/null || true)"
+~/.claude/skills/gstack/bin/gstack-developer-profile --log-session '{"date":"'"$(date -u +%Y-%m-%dT%H:%M:%SZ)"'","mode":"resources","project_slug":"'"${SLUG:-unknown}"'","signal_count":0,"signals":[],"design_doc":"","assignment":"","resources_shown":["URL1","URL2","URL3"],"topics":[]}' 2>/dev/null || true
 ```
 
 2. Log the selection to analytics: