2.8 KiB
name, description
| name | description |
|---|---|
| inspect-shz-logs | Inspect and follow the SHZ backend's local, test, and production logs through bounded, read-only commands. Use when Codex is working in the shz-backend project and needs to diagnose startup failures, exceptions, request behavior, scheduled jobs, or runtime incidents from local backend.log or the test and production backend, error, info, and user logs. |
Inspect SHZ Logs
Use the bundled script instead of assembling SSH commands manually. Keep every initial read bounded; follow logs only when the user explicitly asks to monitor live activity.
Inspect local logs
Read the last 200 lines:
.codex/skills/inspect-shz-logs/scripts/view.sh local 200
Follow the local log interactively:
.codex/skills/inspect-shz-logs/scripts/view.sh local 200 --follow
Inspect production logs
Read the main application log:
.codex/skills/inspect-shz-logs/scripts/view.sh online backend 200
Read the current error log:
.codex/skills/inspect-shz-logs/scripts/view.sh online error 200
online is the existing production-environment name and remains supported. production is an equivalent alias.
Choose only one log type: backend, error, info, or user. Add --follow as the final argument only for an explicit live-monitoring request.
Inspect test logs
Read the main test-environment application log:
.codex/skills/inspect-shz-logs/scripts/view.sh test backend 200
Read the current test-environment error log:
.codex/skills/inspect-shz-logs/scripts/view.sh test error 200
The test environment uses the same four log types as production. The default test server is root@47.111.103.66 on SSH port 22; the endpoint and log directory can be overridden with SHZ_TEST_SSH_HOST, SHZ_TEST_SSH_PORT, and SHZ_TEST_LOG_DIR when the server layout changes.
For keyword analysis, fetch a bounded window and filter locally, for example:
.codex/skills/inspect-shz-logs/scripts/view.sh online error 1000 | rg -i 'exception|error|failed'
Diagnose from logs
- Establish whether the request concerns local, test, or production behavior.
- Read 100–300 recent lines from the relevant log.
- Expand up to the script's 5,000-line cap only when timestamps or stack traces require more context.
- Correlate timestamps, thread names, exception causes, and nearby requests.
- Report the evidence and likely cause without exposing tokens, passwords, personal data, or unrelated log content.
Treat test and production access as read-only. Never restart services, edit files, deploy code, truncate logs, or run arbitrary remote commands through this skill.
Read references/log-locations.md only when checking server identity, paths, or supported log types.