Add prompt hooks

[abhinav-oai]

Why

Codex hooks should support prompt handlers for the events where prompt hooks have defined semantics

What

• Add prompt hook config/API fields: prompt, model, timeout, statusMessage, and continueOnBlock.
• Add a lightweight prompt behavior table with comments for each supported/no-op/unsupported event class.
• Run prompt hooks through the existing hook dispatcher and output parsers by mapping model JSON output into existing hook JSON decisions.
• Wire prompt hook execution to the session model client and regenerate config/app-server schema artifacts.

Example

Tried this hook

[[hooks.UserPromptSubmit.hooks]]
type = "prompt"
statusMessage = "Checking whether the request is actionable"
prompt = "Evaluate if the ask is well specified: $ARGUMENTS. Block requests that are underspecified!"

Results

› help

• UserPromptSubmit hook (blocked)
  feedback: Request is underspecified: clarify what help is needed and the desired outcome.

[abhinav-oai]

https://code.claude.com/docs/en/hooks#response-schema for reference

[abhinav-oai]

[P1] This renders the full hook input JSON into a model request. For PostToolUse and Stop hooks that input can include unbounded tool output or last assistant text, so a single prompt hook can exceed context limits or spend a huge number of tokens before existing hook output spilling applies. Please hard-cap or truncate the rendered prompt before calling the runner.

[abhinav-oai]

since we're not injecting to the main context I don't think its an issue? p sure codex picked this up from the AGENTS.md guidance to limit injected context to 10k toks