Wire up fireHook calls in CLI chat pipeline

[sestinj]

Summary

• Integrates the hooks infrastructure (fireHook.ts) into the CLI chat pipeline so hook events actually fire at runtime
• Previously the hooks system was merged as infrastructure only (PR feat: add hooks system for CLI event interception #11029) with no call sites — hooks never fired
• Wires up all 6 event types: SessionStart, SessionEnd, UserPromptSubmit, PreToolUse, PostToolUse, PostToolUseFailure

Test plan

• Run cn with att hooks configured in ~/.continue/settings.json
• Send a prompt — verify UserPromptSubmit hook fires
• Verify tool use triggers PreToolUse before execution and PostToolUse after
• Exit session — verify SessionEnd hook fires
• Build passes: npm run build in extensions/cli
• Existing tests pass: npm test in extensions/cli

🤖 Generated with Claude Code

---

Continue Tasks: ✅ 7 no changes — View all

---

Summary by cubic

Wires the hooks system into the CLI chat pipeline so events fire during real runs. Enables Claude Code-compatible hooks to observe and control sessions, prompts, and tool calls.

• New Features

  • Registers HookService; loads merged hooks config from Continue/Claude settings with disableAllHooks support.
  • Fires SessionStart, SessionEnd, and UserPromptSubmit (headless and TUI). Runs PreToolUse before permissions (can block or update input). Emits PostToolUse and PostToolUseFailure on success/error.
  • Supports command and HTTP handlers with exit-code semantics, async hooks, regex matchers, env var interpolation, and JSON output for decisions/additionalContext.
  • Adds comprehensive tests, docs, and small CLI wiring updates.

• Bug Fixes

  • Fires SessionStart after services initialize and awaits it in headless mode to prevent early exit.
  • Awaits UserPromptSubmit and cancels the prompt when a hook blocks; headless now prints a JSON error to stderr, and TUI stops active time tracking.
  • Disables eslint max-lines in chat.ts to keep linting passing after integration.

^{Written for commit 44072f1. Summary will update on new commits.}

[cubic-dev-ai]

6 issues found across 14 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="extensions/cli/src/hooks/hooks.test.ts">

<violation number="1" location="extensions/cli/src/hooks/hooks.test.ts:387">
P2: Avoid requiring `python3` in this test command; use a Node-based stdin parser so the test only depends on the project runtime.</violation>
</file>

<file name="extensions/cli/src/hooks/hookRunner.ts">

<violation number="1" location="extensions/cli/src/hooks/hookRunner.ts:164">
P2: Timer resource leak: `clearTimeout` is not called when `fetch()` throws. Move `clearTimeout(timeout)` into a `finally` block to ensure it is always cleaned up.</violation>

<violation number="2" location="extensions/cli/src/hooks/hookRunner.ts:377">
P2: Inconsistent state: a later PreToolUse hook returning `"allow"` overwrites an earlier `"deny"` in `permissionDecision`, but `blocked` remains `true`. Consumers will see contradictory fields. Consider preserving the most restrictive `permissionDecision` (i.e., skip overwrite if already `"deny"`).</violation>
</file>

<file name="extensions/cli/src/commands/chat.ts">

<violation number="1" location="extensions/cli/src/commands/chat.ts:367">
P1: UserPromptSubmit hook result is ignored, so blocking decisions from hooks are not enforced and prompts continue to the model.</violation>

<violation number="2" location="extensions/cli/src/commands/chat.ts:567">
P2: SessionStart hooks can signal a block, but the call isn’t awaited or checked. This makes blocking hooks ineffective during session startup.</violation>

<violation number="3" location="extensions/cli/src/commands/chat.ts:567">
P1: SessionStart hook is fired before hook services are initialized, so the event is dropped and never reaches configured handlers.</violation>
</file>

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

[cubic-dev-ai]

3 issues found across 3 files (changes from recent commits).

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="extensions/cli/src/ui/hooks/useChat.ts">

<violation number="1" location="extensions/cli/src/ui/hooks/useChat.ts:518">
P2: Blocked prompts return before closing telemetry active-time, which can overcount active usage duration.</violation>
</file>

<file name="extensions/cli/src/commands/chat.ts">

<violation number="1" location="extensions/cli/src/commands/chat.ts:369">
P1: Headless blocked prompts are swallowed silently; return path should surface an error so automation can detect policy blocks.</violation>

<violation number="2" location="extensions/cli/src/commands/chat.ts:475">
P2: Await `fireSessionStart` in headless mode so SessionStart hooks reliably complete before potential process exit.</violation>
</file>

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

[cubic-dev-ai]

1 issue found across 2 files (changes from recent commits).

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="extensions/cli/src/commands/chat.ts">

<violation number="1" location="extensions/cli/src/commands/chat.ts:478">
P2: `SessionStart` should be emitted as fire-and-forget, not awaited, to avoid blocking CLI startup.

(Based on your team's feedback about not blocking on SessionStart hooks at startup.) [FEEDBACK_USED]</violation>
</file>

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