Sentry Production Error Explorer
Inspect Sentry issues, events, and basic production health through the Sentry CLI.
The skill explicitly limits itself to read-only queries, requires local authentication, forbids pasting full tokens into chat, and requires PII redaction and no raw stack traces, making the main safety risks visible. Points are deducted because it does not require confirmation for sensitive queries or arbitrary API access, and does not disclose data-transfer scope, service retention, rollback, or dependency-installation risks.
The command examples, defaults, JSON guidance, and authentication flow are reasonably consistent, so the happy path is understandable from static inspection. Points are deducted because there are no tests or reproducible execution artifacts, auto-detection may select the wrong organization or project, and handling of invalid input, permissions, network failures, and unknown resources is thin; static calibration caps this at 10.
The audience and scenarios—Sentry issue/event inspection and basic health queries—are clear, with stated read-only boundaries, input fields, and no-results behavior. Points are deducted because exclusion triggers, supported CLI versions, and Chinese-language output are unspecified; mainland-China network reachability is unsubstantiated, and core operation depends on the external Sentry service.
The documentation has clear sections for quick start, core tasks, inputs/defaults, output rules, and examples. Points are deducted because the repository is marked deprecated, the individual skill license is unknown, and versioning, changelog, maintenance ownership, update path, troubleshooting, and known limitations are missing.
The skill covers issue listing, detail and event retrieval, and basic analysis/plan tasks, with output fields that could be directly usable. Points are deducted because no real output or execution evidence is present, the availability and quality of AI analysis commands are unverified, and the incremental benefit over manual CLI use is limited; static calibration caps this at 7.
The source provides auditable commands, parameters, and behavior claims, and the official organization provenance supports attribution. Points are deducted because there is no committed test suite, CI coverage, third-party execution evidence, or corroborating source; conclusions rely mainly on author documentation, so static calibration caps this at 5.
- The installation instructions pipe a remote curl script to a shell without source verification, version pinning, or cleanup guidance; review it locally before execution.
- Auto-detection from DSNs, environment variables, source code, and directory names can select the wrong target; verify organization and project scope before querying.
- The read-only claim is not demonstrated for every arbitrary API endpoint or AI analysis command; do not assume all operations and data handling have no external effects.
- The repository is marked deprecated and the individual skill license is unknown; confirm the current maintenance source and licensing terms.
- There is no evidence of mainland-China network reachability or Chinese-language output; the core function may be unavailable when the external Sentry service is unreachable.
What it does & when to use it
This is a read-only Sentry observability skill for querying issues, events, and basic health data through the Sentry CLI. It defaults to the previous 24 hours in the production environment and a limit of 20 results, with JSON output for programmatic processing. It can retrieve issue and event details, access arbitrary Sentry API endpoints, and request AI-generated root-cause analyses and fix plans. Users must install and authenticate the Sentry CLI, and outputs should redact personal information and raw stack traces.
Runs sentry issue list to return recent unresolved issues; uses short IDs to retrieve issue details and events; retrieves event details with organization, project, and event ID; runs sentry issue explain for root-cause analysis and sentry issue plan for a fix plan; uses sentry api for uncovered endpoints and sentry schema for endpoint discovery; produces JSON or field-selected output, explicitly reports empty results, and redacts emails, IP addresses, raw stack traces, and authentication tokens.
- A production engineer wants the most recent unresolved production issues from the last 24 hours.
- An on-call responder needs issue details and recent events for a Sentry short ID.
- An engineer needs an event's culprit, timestamp, environment, release, and URL.
- A team wants an AI root-cause analysis or fix plan for a Sentry issue.
- A user needs read-only access to a Sentry API endpoint not covered by a dedicated CLI command.
Pros & cons
- Covers issue lists, issue details, event details, and arbitrary API access.
- Supports JSON and field selection for programmatic workflows.
- Defines read-only behavior and rules for redacting PII, raw stack traces, and tokens.
- The CLI handles authentication, organization/project detection, pagination, and retries.
- Requires the Sentry CLI to be installed and authenticated.
- Requires network access to Sentry and may read local DSNs, environment variables, configuration, or directory names.
- The source provides no test suite, platform verification, or performance data.
- The repository is deprecated, and the source does not provide this skill's license file.
How to install
The skill is located at skills/.curated/sentry/. The README documents installing curated skills in Codex with $skill-installer sentry, followed by a Codex restart. The repository is marked deprecated, and the provided source does not document another installation method for this skill.
How to use
If unauthenticated, run sentry auth login, or set SENTRY_AUTH_TOKEN locally, then confirm with sentry auth status. Example trigger: "List the top 10 open issues for prod in the last 24h." The CLI attempts to detect the organization and project from DSNs, environment variables, configuration defaults, and directory names; pass <org>/<project> explicitly if detection fails or is incorrect. Defaults are a 24-hour period, production environment, and a limit of 20.