fix(types): handle reasoning.encrypted content block type from Grok models

[potb]

Closes #1588

Grok models return reasoning.encrypted content blocks. We only check for "reasoning" and "thinking", so these blocks get silently dropped during validation, extraction, and formatting.

Added reasoning.encrypted to the ThinkingPartType union and THINKING_TYPES set, then updated all downstream checks. Content extraction guards on && part.text so encrypted blocks without text are safely skipped.

[cubic-dev-ai]

3 issues found across 14 files

Confidence score: 4/5

• This PR looks generally safe to merge, but there is a user-facing risk: reasoning.encrypted parts are filtered out before formatting in src/tools/background-task/full-session-format.ts and src/tools/background-task/modules/formatters.ts, so the new handling is unreachable and content gets dropped even when includeThinking is true.
• The test in src/hooks/thinking-block-validator/reasoning-encrypted.test.ts is tautological and doesn’t exercise production logic, so it won’t catch regressions in the validator.
• Pay close attention to src/tools/background-task/full-session-format.ts, src/tools/background-task/modules/formatters.ts, src/hooks/thinking-block-validator/reasoning-encrypted.test.ts - reasoning.encrypted handling is dropped and test coverage is ineffective.

Prompt for AI agents (all issues)

Check if these issues are valid — if so, understand the root cause of each and fix them.


<file name="src/hooks/thinking-block-validator/reasoning-encrypted.test.ts">

<violation number="1" location="src/hooks/thinking-block-validator/reasoning-encrypted.test.ts:5">
P2: Test is tautological: it only checks a locally defined helper instead of exercising production logic, so it provides no real coverage of the thinking-block validator.</violation>
</file>

<file name="src/tools/background-task/full-session-format.ts">

<violation number="1" location="src/tools/background-task/full-session-format.ts:138">
P2: `reasoning.encrypted` parts are filtered out before formatting, so the newly added handling is unreachable and the content is discarded.</violation>
</file>

<file name="src/tools/background-task/modules/formatters.ts">

<violation number="1" location="src/tools/background-task/modules/formatters.ts:305">
P2: `reasoning.encrypted` parts are filtered out before formatting, so the newly added handling is unreachable and these parts are dropped even when `includeThinking` is true.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

[potb]

@cubic-dev-ai recheck

[cubic-dev-ai]

@cubic-dev-ai recheck

@potb I have started the AI code review. It will take a few minutes to complete.

[cubic-dev-ai]

1 issue found across 13 files

Confidence score: 5/5

• Low-severity test quality concern only; no concrete user-facing regression is indicated, so this looks safe to merge.
• src/hooks/thinking-block-validator/reasoning-encrypted.test.ts validates a local helper instead of production logic, so the test could pass even if the real thinking-family check is wrong.
• Pay close attention to src/hooks/thinking-block-validator/reasoning-encrypted.test.ts - test may not exercise actual production behavior.

Prompt for AI agents (all issues)

Check if these issues are valid — if so, understand the root cause of each and fix them.


<file name="src/hooks/thinking-block-validator/reasoning-encrypted.test.ts">

<violation number="1" location="src/hooks/thinking-block-validator/reasoning-encrypted.test.ts:5">
P3: This test defines and validates a local helper instead of exercising production logic, so it will pass even if the real thinking-family check is incorrect (e.g., `startsWithThinkingBlock` still ignores `reasoning.encrypted`). It doesn’t provide regression protection for the intended change.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}