Summary

Removes Enum: []any{true} from the delete property of issue_write's issue_fields item schema, so delete: false is a legal no-op instead of a schema-validation error.

Why

delete was declared {"type": "boolean", "enum": [true]}. The property is optional, but a client that fills every property of a schema — common, since OpenAI-style strict function calling requires every property to appear in required — had no way to say "not deleting this field": there is no false in the enum and no null in the type. The sibling value property offers no alternative either, being typed ["string","number","boolean"] with no null, and issues.go explicitly rejects value: null.

This is not only cosmetic. The MCP Go SDK validates arguments against the resolved input schema before the handler runs (go-sdk@v1.7.0/mcp/server.go, applySchema), so delete: false was rejected at validation and never reached optionalIssueWriteFields. Clients that fill every property instead sent delete: true alongside a value and hit the handler's mutual-exclusion check, so issue_write could never set an issue field for them.

What changed

• pkg/github/issues.go: dropped Enum: []any{true} from the delete property of the issue_fields item schema; extended its description to state that omitting the property, or setting it to false, leaves the field's current value unchanged.
• pkg/github/__toolsnaps__/issue_write.snap: regenerated via UPDATE_TOOLSNAPS=true.
• pkg/github/issues_test.go: added Test_optionalIssueWriteFields, covering delete: false alongside a value, delete: true alone, delete omitted, and both existing error paths. This function previously had no test coverage.

No handler logic changed. optionalIssueWriteFields already reads OptionalParam[bool](itemMap, "delete") and branches on if deleteField, so false falls through to the normal value path, and the delete: true mutual-exclusion check still applies.

MCP impact

• No tool or API changes
• Tool schema or behavior changed
• New tool added

issue_write.issue_fields[].delete widens from enum: [true] to any boolean, and its description gains the no-op case. Strictly a relaxation: every payload valid before is still valid and behaves identically.

Prompts tested (tool changes only)

• Worked against a live GitHub repository with org-level issue fields configured.
• Coverage is the new unit tests over optionalIssueWriteFields plus the regenerated schema snapshot.
• The reproducing payload, from an agent framework against api.githubcopilot.com/mcp/, was {"method": "update", "issue_number": 5, "issue_fields": [{"field_name": "Start date", "value": "2026-08-14", "delete": true, "field_option_name": ""}]} — rejected with issue field "Start date" cannot specify 'delete' together with 'value' or 'field_option_name' on seven consecutive calls across two runs. That client also sent milestone: 0, duplicate_of: 0, and field_option_name: ""; only delete had no benign value. The equivalent prompt is "Set the Start date field on issue #5 to 2026-08-14".
• Caveat: this does not guarantee such a client starts succeeding. Whether a given model emits false rather than true is unverified, and one copying its own prior call will keep sending true. The change removes a schema construct that forces failure on an otherwise recoverable mistake.

Security / limits

• No security or limits impact

Input validation is relaxed for one boolean that routes to an existing code path. No auth, permission, data-exposure, or size-limit surface is touched, and the mutual-exclusion check that prevents delete: true from being combined with value is unchanged.

Tool renaming

• I am renaming tools as part of this PR (e.g. a part of a consolidation effort)
  • I have added the new tool aliases in deprecated_tool_aliases.go
• I am not renaming tools as part of this PR

Lint & tests

• Linted locally with ./script/lint — 0 issues.
• Tested locally with ./script/test — full suite passes, including the snapshot check without UPDATE_TOOLSNAPS set.

Docs

• Not needed
• Updated (README / docs / examples)

Ran ./script/generate-docs; it produced no diff. The generated tables do not include nested item-property descriptions, and a repo-wide grep found the old description only in issue_write.snap.

Fixes #2904.