[sentry-junior]

Part of #1564

Moves the nine allow-listed alert commands off in-func target parsing. Their positionals are now parsed before the command runs.

Rule commands (alert issues view/delete/edit, alert metrics view/delete/edit): these use issueRuleParameter / metricRuleParameter. Both wrap the new alertRuleParameter (src/commands/alert/arguments.ts), which is a boundPositional over the existing parseIssueRuleArg / parseMetricRuleArg splitters. func takes an AlertRuleArgument and reads { ref, target } from await rule.resolved. Rule lookup by ID or name is unchanged. No API calls or endpoints changed.

I didn't add an ID_KINDS entry. Rule names are free text and can contain spaces, while the ID grammar splits IDs at spaces. The rule part also comes after a 1- or 2-segment target, which isn't the [<org>/[<project>/]]<id> shape. The splitters already handle this syntax, so I kept them.

Target-only commands:

• alert issues create uses orgProjectTarget with its resolver, which still requires exactly one project.
• alert metrics create uses orgProjectTarget with resolveOrgOnlyTarget.
• alert metrics list uses listTarget(ORG_PROJECT_TARGET). It keeps its own brief ("project ignored").

Lint and docs:

• Removed the nine alert entries from UNBOUND_TARGET_PARSING_ALLOWED, then the list itself (see below).
• Added AlertRuleArgument to BOUND_TARGET_TYPES so require-target-resolution covers it.
• Added a short "Alert rule positionals" note to packages/cli/AGENTS.md.

Tests:

• argument-conventions.test.ts checks that each rule command parses 42, my-org/web/High error rate and my-org/High error rate the same way.
• It also checks invalid forms (my-org/web for issue rules, my-org/, a/b/c, blank) and target parsing for the create and list commands.
• A pasted URL's server is applied on resolve, not on parse.
• Command tests now set commandPrefix, like Stricli does, and metrics delete spies on resolveOrgOnlyTarget.

Behavior changes

• An invalid rule argument (for example alert metrics view a/b/c, or alert issues view my-org/web without a rule) now fails in Stricli's argument scanner. Before, it failed inside func. The message gets the usual Failed to parse "<input>" for <placeholder>: … prefix and the exit code changes from 21 to 252. The error text is otherwise the same. It also fails before the auth guard now, like other migrated commands.
• A pasted Sentry URL in the target part now has its server applied when the target resolves (after the auth guard), not during parsing.

Lint allow-lists

Drops the last allow-list, per review on #1747; closes out #1564's ratchet. Once #1746 landed, UNBOUND_TARGET_PARSING_ALLOWED held only these nine alert commands. This PR deletes it, along with ratchetedNames and allowedFollowUp. require-target-resolution now uses bannedNames, so it reports every banned reference in command files and has no exemption path. The allow-list and ratchet lint-rule test cases are removed, and packages/cli/AGENTS.md now says neither positional rule has an allow-list.

Not in scope: alert rule URLs (/alerts/rules/..., /monitors/...) were not accepted before and still aren't. sentry-url-parser has no alert-rule paths. That could be a follow-up.

Verification: tsc, pnpm run lint, lint-rule tests (scripts/cli-lint-rules.test.mjs, 69/69 after the cleanup) and the alert and argument-conventions tests pass. Before the merge, the full CLI unit suite had 12593 passing tests and only 6 failures, all known root-permission sandbox tests (EACCES/readonly).

via BYK.

--

View Junior Session in Sentry

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
sentry-local	Error		Oct 29, 2026 11:55pm UTC