@m13v/s4l 1.7.1-rc.2 → 1.7.1-rc.21

package/skill/sentry-digest.sh

@@ -1,19 +1,132 @@
 #!/usr/bin/env bash
 # sentry-digest.sh — critical (level:error/fatal) Sentry issue digest for the
-# s4l Sentry project. Replaces the raw per-issue Sentry alert email; only
-# emails when something is NEW or GROWING. Idempotent via
-# scripts/state/sentry_digest_ledger.json. Wired by
-# launchd/com.m13v.s4l-sentry-digest.plist (every 4 hours).
+# s4l Sentry project. Replaces the raw per-issue Sentry alert email.
+#
+# Two-step design (mirrors ~/fazm/inbox/skill/check-query-insights.sh):
+#   1. scripts/sentry_digest.py pulls issues + diffs against the ledger
+#      (mechanical, no LLM). If nothing is NEW or GROWING, exit here, no
+#      Claude spawn, no email, no ledger write (ledger is only ever written
+#      by the investigating Claude session, see SENTRY-DIGEST-SKILL.md).
+#   2. If something changed, spawn Claude to actually investigate (read the
+#      crashing code, check git log, judge actionability) and send ONE
+#      human-readable email, not a raw data dump.
+#
+# Wired by launchd/com.m13v.s4l-sentry-digest.plist (every 4 hours).
 
-REPO_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+set -euo pipefail
+
+# claude (claude-account-rotator wrapper) and gtimeout must be resolvable
+# regardless of how this script is invoked; launchd's plist PATH covers this,
+# but keep it explicit so a manual run also works.
+export PATH="$HOME/.nvm/versions/node/v20.19.4/bin:/opt/homebrew/bin:/usr/local/bin:$HOME/claude-account-rotator/bin:$PATH"
+
+source "$(dirname "$0")/lock.sh"
+acquire_lock "sentry-digest" 1800
+
+REPO_DIR="$HOME/social-autoposter"
+SCRIPTS_DIR="$REPO_DIR/scripts"
 LOG_DIR="$REPO_DIR/skill/logs"
+LEDGER_PATH="$REPO_DIR/scripts/state/sentry_digest_ledger.json"
+CLAUDE_BIN="${CLAUDE_BIN:-$HOME/claude-account-rotator/bin/claude}"
+PYTHON_BIN="${PYTHON_BIN:-/opt/homebrew/bin/python3.11}"
+[ -x "$PYTHON_BIN" ] || PYTHON_BIN="/usr/bin/python3"
 mkdir -p "$LOG_DIR"
-LOG_FILE="$LOG_DIR/sentry-digest-$(date +%Y%m%d).log"
 
-cd "$REPO_DIR" || exit 1
+TS=$(date +%Y-%m-%d_%H%M%S)
+LOG_FILE="$LOG_DIR/sentry-digest-$TS.log"
+SCAN_FILE="$LOG_DIR/sentry-digest-scan-$TS.json"
+OUTCOME_FILE="$LOG_DIR/outcome-sentry-digest-$TS.json"
+
+log() { echo "[$(date +%H:%M:%S)] $*" | tee -a "$LOG_FILE"; }
+
+cd "$REPO_DIR"
+
+log "=== Sentry digest scan: $(date) ==="
+"$PYTHON_BIN" "$SCRIPTS_DIR/sentry_digest.py" --out "$SCAN_FILE" 2>&1 | tee -a "$LOG_FILE"
+SCAN_EXIT=${PIPESTATUS[0]}
+if [ "$SCAN_EXIT" -ne 0 ]; then
+    log "ERROR: sentry_digest.py scan failed (exit $SCAN_EXIT)"
+    exit 1
+fi
+
+FIRST_RUN=$("$PYTHON_BIN" -c "import json; print(json.load(open('$SCAN_FILE'))['firstRun'])")
+NEW_COUNT=$("$PYTHON_BIN" -c "import json; print(json.load(open('$SCAN_FILE'))['newCount'])")
+GROWING_COUNT=$("$PYTHON_BIN" -c "import json; print(json.load(open('$SCAN_FILE'))['growingCount'])")
+
+if [ "$FIRST_RUN" = "False" ] && [ "$NEW_COUNT" -eq 0 ] && [ "$GROWING_COUNT" -eq 0 ]; then
+    log "Nothing new or growing. Skipping Claude spawn and email."
+    find "$LOG_DIR" -name "sentry-digest-scan-*.json" -mtime +7 -delete 2>/dev/null || true
+    find "$LOG_DIR" -name "sentry-digest-*.log" -mtime +30 -delete 2>/dev/null || true
+    exit 0
+fi
+
+# Step 2: spawn Claude to investigate and email
+PROMPT_FILE=$(mktemp)
+cat > "$PROMPT_FILE" <<PROMPT_EOF
+Read ~/social-autoposter/skill/SENTRY-DIGEST-SKILL.md for the full workflow.
+
+## Run metadata
+
+- First run (ledger was empty): $FIRST_RUN
+- New issues this scan: $NEW_COUNT
+- Growing issues this scan: $GROWING_COUNT
+- Sentry auth token: read it yourself with
+  security find-generic-password -s sentry-auth-token -w
+
+## Scan result (from sentry_digest.py)
+
+Full result is on disk at:
+$SCAN_FILE
+
+## Ledger path
+
+$LEDGER_PATH
+
+## Outcome file
+
+Write your JSON outcome to:
+$OUTCOME_FILE
+
+Now run the workflow.
+PROMPT_EOF
+
+log "Spawning Claude Code session for investigation..."
+log "  Scan file: $SCAN_FILE"
+log "  Outcome file: $OUTCOME_FILE"
+
+CLAUDE_EXIT=0
+gtimeout 1800 "$CLAUDE_BIN" \
+    -p "$(cat "$PROMPT_FILE")" \
+    --dangerously-skip-permissions \
+    2>&1 | tee -a "$LOG_FILE" || CLAUDE_EXIT=$?
+
+rm -f "$PROMPT_FILE"
+
+if [ $CLAUDE_EXIT -eq 124 ]; then
+    log "ERROR: Claude Code timed out after 30 minutes"
+elif [ $CLAUDE_EXIT -ne 0 ]; then
+    log "WARNING: Claude Code exited with code $CLAUDE_EXIT"
+fi
+
+log "--- Post-run validation ---"
+if [ -f "$OUTCOME_FILE" ]; then
+    log "Outcome file found:"
+    cat "$OUTCOME_FILE" >> "$LOG_FILE"
+    REPORT_SENT=$("$PYTHON_BIN" -c "import json; d=json.load(open('$OUTCOME_FILE')); print(d.get('reportEmailSent', False))" 2>/dev/null || echo "False")
+    LEDGER_WRITTEN=$("$PYTHON_BIN" -c "import json; d=json.load(open('$OUTCOME_FILE')); print(d.get('ledgerWritten', False))" 2>/dev/null || echo "False")
+    SUMMARY=$("$PYTHON_BIN" -c "import json; d=json.load(open('$OUTCOME_FILE')); print(d.get('summary', 'No summary'))" 2>/dev/null || echo "No summary")
+    log "  reportEmailSent: $REPORT_SENT"
+    log "  ledgerWritten: $LEDGER_WRITTEN"
+    log "  summary: $SUMMARY"
+    if [ "$LEDGER_WRITTEN" != "True" ]; then
+        log "WARNING: agent did not confirm writing the ledger; next run may re-flag the same issues"
+    fi
+else
+    log "WARNING: No outcome file produced. Agent may not have completed the workflow."
+fi
+
+log "=== Done Sentry digest run (claude_exit=$CLAUDE_EXIT) ==="
 
-{
-    echo "=== $(date -u +%Y-%m-%dT%H:%M:%SZ) sentry-digest run ==="
-    /usr/bin/python3 scripts/sentry_digest.py
-    echo
-} >> "$LOG_FILE" 2>&1
+find "$LOG_DIR" -name "sentry-digest-*.log" -mtime +30 -delete 2>/dev/null || true
+find "$LOG_DIR" -name "sentry-digest-scan-*.json" -mtime +7 -delete 2>/dev/null || true
+find "$LOG_DIR" -name "outcome-sentry-digest-*.json" -mtime +30 -delete 2>/dev/null || true