Skip to content

Business operating system for Claude Code — 57 skills, 21 agents, smart daemon. Unified inbox (WhatsApp/Email/Slack/Telegram), autonomous PR merge, full-AWS monitoring, revenue (Stripe+RevenueCat), e-commerce (Shopify), marketing (Klaviyo/Meta/GA4), voice (Bland/ElevenLabs), APM (Datadog/NewRelic/OTEL), YOLO mode.

fix(fires): pre-gather Sentry, want de MCP-toolnaam bestond niet (#942)

Aurora Capital06ee5fa42a29
diff --git a/claude-ops/CHANGELOG.md b/claude-ops/CHANGELOG.md
index eb0af0f2..6c7e57d8 100644
--- a/claude-ops/CHANGELOG.md
+++ b/claude-ops/CHANGELOG.md
@@ -1,5 +1,35 @@
## Unreleased
+- fix(fires): `/ops:fires` reported Sentry as unavailable on every run. The
+ `allowed-tools` frontmatter named `mcp__sentry__*`, but the plugin registers
+ the tools as `mcp__plugin_sentry_sentry__*`, so the tool never existed under
+ the name the skill asked for and the skill dutifully said "unavailable".
+ Sentry is now pre-gathered via `bin/ops-sentry` (REST), the same way infra, CI
+ and external health already are, so it no longer depends on the model
+ remembering to call a tool. The tool names are corrected in `ops-fires`,
+ `ops-triage`, `ops-orchestrate`, `build-fix` and `deploy-fix` for the paths
+ that still use MCP (stack traces, Seer). A non-null `error` from the probe now
+ reads as a gap in coverage, never as "production is clean". The script never
+ exits non-zero and never prints anything but valid JSON — the skill inlines
+ its output, and one stray error line would corrupt the prompt. It also honours
+ the org's region host: Sentry is region-sharded and the wrong region returns
+ an empty list rather than an error.
+
+ Found the hard way: a fires run showed 22 CI failures (19 of them stale cache,
+ already green) and zero Sentry issues, while `ServiceUnavailableException` on
+ Bedrock `Converse` with `reached max retries: 0` had been affecting 4 users
+ for thirteen hours. The only fire with real user impact was the one missing
+ from the board.
+
+- fix(projects): `ops-projects` rendered `None` for any field whose JSON value
+ was an explicit `null`. `.get(key, default)` returns the `null`, not the
+ default; now `or`.
+
+- fix(daemon): leave a deliberate daemon wrapper in the plugin data dir alone
+ instead of rewriting the plist back to `$PLUGIN_ROOT`. The data-dir plist
+ guard wrote the wrapper path straight back, and each write booted the daemon
+ out before it finished its first monitor pass.
+
- fix(hooks): UserPromptSubmit inbox autosync no longer interpolates `$INPUT`.
Grok treats `$VAR` in a hook command as a required env var and skips the
hook when it is unset. The script already reads the event JSON from stdin.
diff --git a/claude-ops/bin/ops-projects b/claude-ops/bin/ops-projects
index 516a9c2a..66906eca 100755
--- a/claude-ops/bin/ops-projects
+++ b/claude-ops/bin/ops-projects
@@ -116,14 +116,14 @@ for p in projects:
idle.append(p)
def fmt_proj(p, indent=" "):
- name = p.get("name","?")
- phase = p.get("phase","—")
- status = p.get("status","—")
- branch = p.get("branch","—")
- dirty = p.get("uncommitted",0)
+ name = p.get("name") or "?"
+ phase = p.get("phase") or "—"
+ status = p.get("status") or "—"
+ branch = p.get("branch") or "—"
+ dirty = p.get("uncommitted") or 0
source = p.get("source","")
- blockers = p.get("blockers","0")
- next_action = p.get("next_action","")[:40] or "—"
+ blockers = p.get("blockers") or "0"
+ next_action = (p.get("next_action") or "")[:40] or "—"
short_status = status[:30] if status else "—"
return f"{indent}{name:<22} {phase:<8} {short_status:<32} {branch:<12} {dirty} dirty {next_action}"
diff --git a/claude-ops/bin/ops-sentry b/claude-ops/bin/ops-sentry
new file mode 100755
index 00000000..dbdb578e
--- /dev/null
+++ b/claude-ops/bin/ops-sentry
@@ -0,0 +1,117 @@
+#!/usr/bin/env bash
+# ops-sentry — Unresolved Sentry issues for the FIRES dashboard → JSON
+#
+# Why this exists: /ops:fires used to declare the Sentry MCP in allowed-tools
+# and hope the model would call it. It never did reliably — the tool name in
+# the frontmatter (mcp__sentry__*) does not match the name the plugin actually
+# registers (mcp__plugin_sentry_sentry__*), so the tool was never available and
+# every run reported Sentry as "unavailable". Pre-gathering it via the REST API
+# puts Sentry on the same footing as infra, CI and external projects: the data
+# is simply in the prompt, with nothing left to remember.
+#
+# Emits {"org":..., "issues":[...], "error":...}. NEVER exits non-zero and
+# never emits anything but valid JSON — the skill inlines this and a bare
+# error string would corrupt the prompt.
+#
+# Token: env SENTRY_AUTH_TOKEN → env SENTRY_TOKEN → doppler claude-ops/prd
+# Org: $1 → preferences.json .partner_registry.sentry.org → auto-discover
+# Region: preferences.json .partner_registry.sentry.region_url → us.sentry.io
+# (Sentry is region-sharded; the org's regionUrl is authoritative and
+# a wrong region returns an empty list rather than an error.)
+
+set -uo pipefail # deliberately no -e: a failed probe must degrade, not abort
+
+API_VERSION="0"
+PERIOD="${OPS_SENTRY_PERIOD:-24h}"
+LIMIT="${OPS_SENTRY_LIMIT:-10}"
+CURL_TIMEOUT="${OPS_SENTRY_TIMEOUT:-12}"
+DEFAULT_REGION="https://us.sentry.io"
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+OPS_PLUGIN_ROOT_FALLBACK="${SCRIPT_DIR}/.." . "${SCRIPT_DIR}/../lib/registry-path.sh" 2>/dev/null || true
+PREFS="${OPS_DATA_DIR:-$HOME/.claude/plugins/data/ops-ops-marketplace}/preferences.json"
+
+# Bail out with valid JSON and exit 0, always.
+emit_error() {
+ jq -nc --arg e "$1" '{org:null, issues:[], error:$e}'
+ exit 0
+}
+
+command -v jq >/dev/null 2>&1 || { printf '{"org":null,"issues":[],"error":"jq not installed"}\n'; exit 0; }
+command -v curl >/dev/null 2>&1 || emit_error "curl not installed"
+
+pref() {
+ [ -f "$PREFS" ] || return 0
+ jq -r "$1 // empty" "$PREFS" 2>/dev/null || true
+}
+
+# ---- token -----------------------------------------------------------------
+TOKEN="${SENTRY_AUTH_TOKEN:-${SENTRY_TOKEN:-}}"
+if [ -z "$TOKEN" ] && command -v doppler >/dev/null 2>&1; then
+ TOKEN=$(doppler secrets get SENTRY_AUTH_TOKEN --project claude-ops --config prd --plain 2>/dev/null || true)
+fi
+[ -z "$TOKEN" ] && emit_error "no Sentry token (env SENTRY_AUTH_TOKEN / SENTRY_TOKEN / doppler claude-ops/prd)"
+
+REGION_URL="$(pref '.partner_registry.sentry.region_url')"
+[ -z "$REGION_URL" ] && REGION_URL="$DEFAULT_REGION"
+REGION_URL="${REGION_URL%/}"
+
+api() {
+ curl -sS --max-time "$CURL_TIMEOUT" \
+ -H "Authorization: Bearer ${TOKEN}" \
+ -H "Accept: application/json" \
+ "$@" 2>/dev/null
+}
+
+# ---- org -------------------------------------------------------------------
+ORG="${1:-}"
+[ -z "$ORG" ] && ORG="$(pref '.partner_registry.sentry.org')"
+if [ -z "$ORG" ]; then
+ # Org listing is not region-sharded; the control silo answers for every org.
+ ORG_RAW=$(api "https://sentry.io/api/${API_VERSION}/organizations/")
+ if printf '%s' "$ORG_RAW" | jq -e 'type=="array"' >/dev/null 2>&1; then
+ ORG=$(printf '%s' "$ORG_RAW" | jq -r '.[0].slug // empty' 2>/dev/null || true)
+ else
+ # An object here means the token was rejected. Say so, rather than blaming
+ # the org — a bad token and a missing org need different fixes.
+ AUTH_DETAIL=$(printf '%s' "$ORG_RAW" | jq -r '.detail // .error // empty' 2>/dev/null || true)
+ [ -n "$AUTH_DETAIL" ] && emit_error "Sentry auth failed: ${AUTH_DETAIL}"
+ fi
+fi
+[ -z "$ORG" ] && emit_error "could not resolve a Sentry org (set .partner_registry.sentry.org in preferences.json)"
+
+# ---- issues ----------------------------------------------------------------
+# statsPeriod bounds the window; query filters to still-open issues.
+RAW=$(api -G "${REGION_URL}/api/${API_VERSION}/organizations/${ORG}/issues/" \
+ --data-urlencode "query=is:unresolved" \
+ --data-urlencode "statsPeriod=${PERIOD}" \
+ --data-urlencode "limit=${LIMIT}" \
+ --data-urlencode "sort=freq")
+
+[ -z "$RAW" ] && emit_error "Sentry API returned no response for org ${ORG} (region ${REGION_URL})"
+
+# An error body is a JSON object; a good body is an array. Distinguish, so an
+# expired token reads as "expired token" and not as "zero issues".
+if ! printf '%s' "$RAW" | jq -e 'type=="array"' >/dev/null 2>&1; then
+ DETAIL=$(printf '%s' "$RAW" | jq -r '.detail // .error // "unrecognised response"' 2>/dev/null || echo "unparseable response")
+ jq -nc --arg o "$ORG" --arg e "Sentry API error: $DETAIL" '{org:$o, issues:[], error:$e}'
+ exit 0
+fi
+
+printf '%s' "$RAW" | jq -c --arg o "$ORG" --arg p "$PERIOD" '{
+ org: $o,
+ period: $p,
+ issues: [ .[] | {
+ short_id: .shortId,
+ title: .title,
+ culprit: .culprit,
+ project: (.project.slug // null),
+ level: .level,
+ events: (.count // "0"),
+ users: (.userCount // 0),
+ first_seen: .firstSeen,
+ last_seen: .lastSeen,
+ url: .permalink
+ } ],
+ error: null
+}' 2>/dev/null || emit_error "failed to parse Sentry issue list"
diff --git a/claude-ops/prompts/build-fix.md b/claude-ops/prompts/build-fix.md
index 8242eabd..f935b0b3 100644
--- a/claude-ops/prompts/build-fix.md
+++ b/claude-ops/prompts/build-fix.md
@@ -11,7 +11,7 @@ You are headless inside Claude Code with full claude-ops tooling:
| Mobile / Expo / RN / iOS Fastlane | spawn the `Mobile App Specialist` subagent |
| TypeScript / type errors | spawn the `typescript-reviewer` subagent |
| Cross-repo contract breakage | spawn the `fullstack-mobile-architect` subagent |
-| Sentry context for runtime crashes | `mcp__sentry__search_events` |
+| Sentry context for runtime crashes | `mcp__plugin_sentry_sentry__search_events` |
| Apple Developer / TestFlight state | `python3 scripts/asc-manage-builds.py` (in your mobile repo) |
| Doppler secret resolution | `doppler secrets get <KEY> --project <your-project> --config <env> --plain` |
| Library version drift | `npx expo install --check`, `npm ls <pkg>`, Context7 MCP for docs |
diff --git a/claude-ops/prompts/deploy-fix.md b/claude-ops/prompts/deploy-fix.md
index 593f819a..a1285a73 100644
--- a/claude-ops/prompts/deploy-fix.md
+++ b/claude-ops/prompts/deploy-fix.md
@@ -10,7 +10,7 @@ You are running headless inside a Claude Code session with full claude-ops tooli
| -------------------------------------------- | ------------------------------------------------------------ |
| Fetch failed CI logs | `gh run view {{RUN_ID}} --repo {{REPO}} --log-failed` |
| Inspect ECS / AWS state | `/ops:ops-fires`, `/ops:ops-monitor`, raw `aws` CLI |
-| Sentry context for related errors | `/ops:ops-triage` or `mcp__sentry__search_events` |
+| Sentry context for related errors | `/ops:ops-triage` or `mcp__plugin_sentry_sentry__search_events` |
| Find prior fixes for similar failures | `gh search prs --repo {{REPO}} 'fix(deploy)' --state merged` |
| My-Project mobile crashes / iOS build issues | spawn the `Mobile App Specialist` subagent |
| My-Project backend / NestJS issues | spawn the `Health Data Expert` subagent |
diff --git a/claude-ops/skills/ops-fires/SKILL.md b/claude-ops/skills/ops-fires/SKILL.md
index 77d4c6d0..7e25484f 100644
--- a/claude-ops/skills/ops-fires/SKILL.md
+++ b/claude-ops/skills/ops-fires/SKILL.md
@@ -17,8 +17,9 @@ allowed-tools:
- Monitor
- WebFetch
- WebSearch
- - mcp__sentry__search_issues
- - mcp__sentry__get_issue_details
+ - mcp__plugin_sentry_sentry__search_issues
+ - mcp__plugin_sentry_sentry__get_sentry_resource
+ - mcp__plugin_sentry_sentry__find_organizations
effort: medium
maxTurns: 30
---
@@ -41,7 +42,8 @@ Before executing, load available context:
- First: check `$AWS_ACCESS_KEY_ID` / `$AWS_PROFILE` env vars
- Then: `doppler secrets get AWS_ACCESS_KEY_ID --plain` (if `doppler` configured in prefs)
- Then: use `password_manager_config.query_cmd` from preferences
- - Sentry token: `$SENTRY_AUTH_TOKEN` → Doppler `SENTRY_AUTH_TOKEN` → vault
+ - Sentry token: `$SENTRY_AUTH_TOKEN` → `$SENTRY_TOKEN` → Doppler `claude-ops/prd/SENTRY_AUTH_TOKEN`. Resolved by `bin/ops-sentry`; no action needed here.
+ - Sentry org and region: `preferences.json` `.partner_registry.sentry.{org,region_url}`. Sentry is region-sharded — issues are served by the org's own region host (e.g. `https://us.sentry.io`), not `sentry.io`, and querying the wrong region returns an empty list rather than an error. Omit `org` and the script discovers the first org the token can see.
3. **Preferences**: Read `${CLAUDE_PLUGIN_DATA_DIR}/preferences.json` for `secrets_manager` config to know which vault to query.
@@ -67,7 +69,9 @@ Before executing, load available context:
| Command | Usage | Output |
| -------------------------------------------------------------------------------------------------------------------------------- | --------------------------------- | ---------- |
| `sentry-cli issues list --project <slug> --status unresolved` | Unresolved issues | Issue list |
-| `curl -H "Authorization: Bearer $SENTRY_AUTH_TOKEN" "https://sentry.io/api/0/projects/<org>/<proj>/issues/?query=is:unresolved"` | API fallback when MCP unavailable | JSON array |
+| `bin/ops-sentry [org]` | Pre-gather path (already inlined below) | JSON `{org,issues,error}` |
+| `OPS_SENTRY_PERIOD=7d OPS_SENTRY_LIMIT=25 bin/ops-sentry` | Widen the window or cap | JSON |
+| `curl -H "Authorization: Bearer $SENTRY_AUTH_TOKEN" "https://us.sentry.io/api/0/organizations/<org>/issues/?query=is:unresolved"` | Manual probe (note: region host, org-scoped) | JSON array |
---
@@ -111,6 +115,17 @@ ${CLAUDE_PLUGIN_ROOT}/scripts/finops-bridge.sh anomalies high 2>/dev/null || ech
${CLAUDE_PLUGIN_ROOT}/bin/ops-ci 2>/dev/null || echo '[]'
```
+## Sentry — unresolved issues (last 24h)
+
+Pre-gathered so Sentry is never skipped. Sorted by event frequency. An
+`error` field that is not null means the probe itself failed (missing or
+expired token, unreachable API) — report that as a gap, and do NOT read an
+empty `issues` list as "no errors in production".
+
+```!
+${CLAUDE_PLUGIN_ROOT}/bin/ops-sentry 2>/dev/null || echo '{"org":null,"issues":[],"error":"sentry probe failed"}'
+```
+
## External projects health
```!
@@ -132,7 +147,7 @@ fi
Analyze the pre-gathered data — including external projects. Then run parallel checks:
1. **ECS health** — parse infra data for unhealthy services, stopped tasks, failed deployments.
-2. **Sentry** — if Sentry MCP is connected, query recent unresolved errors. Otherwise note it's unavailable.
+2. **Sentry** — parse the pre-gathered Sentry data above; it is always present, so never report Sentry as unchecked. Classify by blast radius: an issue affecting many users or growing fast is HIGH, a warning-level issue with a handful of events is LOW. Only reach for `mcp__plugin_sentry_sentry__get_sentry_resource` (stack trace, breadcrumbs) or `analyze_issue_with_seer` when you are about to dispatch a fix agent for that issue. If the probe returned a non-null `error`, say Sentry could not be read and why — an empty list is not proof production is clean.
3. **CI** — parse CI data for failing pipelines, broken main/dev branches.
4. **GitHub Actions** — `gh run list --limit 20 --json status,conclusion,name,headBranch,createdAt 2>/dev/null`
5. **External projects** — parse ops-external data. Flag `auth_expired` as HIGH (credential rotation needed), `unreachable`/`degraded` as MEDIUM, `not_configured` as LOW.
@@ -175,8 +190,9 @@ ECS HEALTH
CI STATUS
[repo] [branch] [workflow] [status] [last run]
-SENTRY (top errors, 24h)
-[error] [count] [first seen] [project]
+SENTRY (unresolved, 24h — by event count)
+[short_id] [title truncated] [events]ev [users]u [project] [last seen]
+[If the probe returned an error, print that instead of an empty section]
EXTERNAL PROJECTS
[alias] [source] [status] [details — e.g. auth_expired, unreachable]
diff --git a/claude-ops/skills/ops-orchestrate/SKILL.md b/claude-ops/skills/ops-orchestrate/SKILL.md
index 05022336..7b059815 100644
--- a/claude-ops/skills/ops-orchestrate/SKILL.md
+++ b/claude-ops/skills/ops-orchestrate/SKILL.md
@@ -27,7 +27,7 @@ allowed-tools:
- CronCreate
- CronList
- LSP
- - mcp__sentry__search_issues
+ - mcp__plugin_sentry_sentry__search_issues
- mcp__linear__list_issues
- mcp__linear__update_issue
- mcp__linear__create_issue
@@ -178,7 +178,7 @@ gh run list --repo <repo> --workflow "<workflow>" --limit 5 --json conclusion,he
Run these in parallel with the audit agents:
-- **Sentry**: `mcp__sentry__search_issues` or `sentry-cli issues list` — P0/P1 unresolved errors
+- **Sentry**: `mcp__plugin_sentry_sentry__search_issues` or `sentry-cli issues list` — P0/P1 unresolved errors
- **Linear**: `mcp__linear__list_issues` for current sprint — in-progress and unstarted
- **GitHub Issues**: `gh issue list --state open` per repo
- **Conversation context**: Scan current conversation for todos, requests, and incomplete work the user mentioned
diff --git a/claude-ops/skills/ops-triage/SKILL.md b/claude-ops/skills/ops-triage/SKILL.md
index 6f3dbec1..fc17ff6a 100644
--- a/claude-ops/skills/ops-triage/SKILL.md
+++ b/claude-ops/skills/ops-triage/SKILL.md
@@ -18,8 +18,8 @@ allowed-tools:
- WebFetch
- WebSearch
- LSP
- - mcp__sentry__search_issues
- - mcp__sentry__get_issue_details
+ - mcp__plugin_sentry_sentry__search_issues
+ - mcp__plugin_sentry_sentry__get_sentry_resource
- mcp__linear__list_issues
- mcp__linear__update_issue
- mcp__linear__get_issue