What problem does it solve?
It standardizes how user-facing agents turn raw conversation into importable workflow requests, preventing ad-hoc implementation, duplicated submissions, and unvalidated status reporting in a multi-agent CCB pipeline.
Core Features & Use Cases
- Strict message classification: Every user turn is classified as a direct answer, macro task request, blocked prerequisite, or final report before any action is taken.
- Structured intake artifacts: Produces exact
**Intake Evidence** or **Blocked Evidence** markdown shapes with a unique CCB_REQ_ID, scope, required behavior, and constraints for the Planner.
- Single controlled handoff: Submits evidence exactly once via the Codex
ccb_frontdesk_ask_planner tool or the Claude silent ask command, then stops.
- Validated final reporting: Renders only Planner-authored
user_report_body from a validated ccb.planner.frontdesk_status.v1 envelope without reconstructing reports from child logs.
- Use Case: A user asks the frontdesk agent to create a documentation file; the skill returns an Intake Evidence artifact naming the path, required content, and authority limits, then hands off to the Planner instead of creating the file.
Quick Start
Ask the frontdesk agent to handle a new project request and it will classify the message, emit a structured Intake Evidence artifact, and forward it to the Planner exactly once.
