AI 原生开放教材

外观

The Context Course系统课程·英文原文 ↗原件 ↓仓库 ↗

Quiz 1: Hook Fundamentals ​

Test your understanding of hooks, the events they cover, and how they differ across Claude Code, Codex, OpenCode, and Pi.

Question 1: What is a hook? ​

选项

  • A. A hook is a model prompt that reminds the agent to run a command
  • B. A hook is a user-defined handler that runs at a deterministic point in the agent's lifecycle, independent of the model's cooperation
  • C. A hook is a way to package skills, MCP servers, and agents into a distributable bundle
  • D. A hook is a kind of subagent that runs in parallel with the main agent

答案解析

  • A — That's a skill, not a hook. Hooks are runtime handlers — the runtime calls them regardless of what the model chose to do.
  • B（正确答案） — Correct! Hooks fire on events like PreToolUse and Stop, and they run whether or not the model 'wanted' them to.
  • C — That's a plugin. Some plugin systems (like Claude Code's) can bundle hooks, but a hook is not the same as a plugin.
  • D — Subagents are isolated model instances. Hooks are runtime callbacks — much lighter weight and not model calls by default.

Question 2: How do hook surfaces differ across platforms? ​

选项

  • A. Claude Code uses PascalCase JSON events (PreToolUse, Stop); Codex uses a smaller PascalCase JSON set behind a feature flag; OpenCode uses TS/JS plugin modules; Pi uses TS/JS extensions with lower_snake_case events like tool_call
  • B. All four platforms use the exact same settings.json format for hook events
  • C. OpenCode exposes hooks through JSON configuration, like Claude Code
  • D. Only Claude Code supports HTTP hooks; Codex, OpenCode, and Pi cannot send events to an HTTP receiver

答案解析

  • A（正确答案） — Correct! The four platforms share the same lifecycle but differ in syntax and surface.
  • B — False. Only Claude Code uses settings.json. Codex uses hooks.json, OpenCode uses code-first plugins, and Pi uses extensions.
  • C — False. OpenCode has no JSON event config — plugins are TS/JS modules exporting hook handlers as object keys.
  • D — Close but not quite. Claude Code's 'http' handler type is unique, but Codex command hooks can call curl, and OpenCode and Pi extensions can call fetch() — all can deliver events to HTTP.

Question 3: Which lifecycle events matter most? ​

选项

  • A. PreToolUse fires after the tool returns, so you can log the result
  • B. UserPromptSubmit fires when the user submits a prompt, before the model sees it — ideal for injecting extra context
  • C. Stop fires only at session end, not per turn
  • D. PostToolUse in Codex fires for every possible tool path, just like in Claude Code

答案解析

  • A — False. PreToolUse fires before the tool call. PostToolUse is the 'after' event.
  • B（正确答案） — Correct! This is the right event for prepending or modifying the prompt. Claude Code and Codex support UserPromptSubmit directly; Pi's nearest equivalent is before_agent_start.
  • C — False. Stop fires at the end of each turn; SessionEnd (Claude Code) is the session-level event.
  • D — False. Codex PreToolUse and PostToolUse cover supported tool calls such as Bash, apply_patch, and MCP tools, but the surface is narrower than Claude Code's.

Question 4: How do hooks block or modify actions? ​

选项

  • A. A Claude Code command hook exits with code 2 and writes a reason to stderr to block the action
  • B. A Claude Code hook returns HTTP 500 to block the action
  • C. An OpenCode hook writes to stdin to block a tool call
  • D. Hooks cannot block or modify tool calls; they are observe-only

答案解析

  • A（正确答案） — Correct! Exit 2 + stderr is the documented blocking signal; on PreToolUse it blocks the tool call, on UserPromptSubmit it erases the prompt.
  • B — False. Non-2xx responses from HTTP hooks are treated as non-blocking errors, not as explicit block signals.
  • C — False. OpenCode hooks block by throwing an Error. There is no stdin — they are TS/JS functions, not processes.
  • D — False. All four platforms support influencing the agent: Claude Code and Codex via exit codes or JSON output, OpenCode via mutating output or throwing, and Pi via returned values or event mutation.

Question 5: When should you use hooks instead of skills? ​

选项

  • A. Hooks replace skills: once you have hooks, you don't need to write skills anymore
  • B. Hooks are the right surface when you need deterministic behavior — like 'always run the linter after an Edit' — that shouldn't depend on the model remembering
  • C. Hooks are mainly for user interface theming
  • D. Hooks only make sense for single-agent workflows

答案解析

  • A — False. Skills teach the model when and how to do something; hooks fire around every step regardless of what the model chose. They complement each other.
  • B（正确答案） — Correct! That's exactly what hooks are for: turning conventions into code so the runtime enforces them.
  • C — False. A few events (like Claude Code's Notification) relate to UI, but hooks are mostly about observability, guardrails, and automation.
  • D — False. SubagentStart / SubagentStop (Claude Code) and session.* events (OpenCode) make hooks useful in multi-agent workflows too.

Summary ​

If you got 4-5 correct, you understand the core hook model well enough to start wiring real guardrails and observability. If not, review the event lifecycle and the platform comparison before continuing.

Key Takeaways ​

  • Hooks are deterministic runtime handlers, not prompts or plugins
  • Claude Code, Codex, OpenCode, and Pi expose similar lifecycle moments with different configuration surfaces
  • Hooks are the right tool when behavior must happen every time rather than relying on the model to remember

Next Steps ​

Next, wire hooks into a real dashboard so you can see agent activity live.