dcr-lead-init

name	dcr-lead-init
description	Start Deep Code Review as the lead for an active review run.
invocation-args	{"task_name":"required"}

dcr-lead-init

Do not start this playbook implicitly.

Bind the lead run to an explicit task_name at invocation time.

Read First

1. .agent/skills/code-review/SKILL.md
2. .agent/playbooks/deep-code-review/PLAYBOOK.md
3. .agent/playbooks/deep-code-review/transport.md
4. .agent/playbooks/deep-code-review/template/round_template/README.md

Keep these hot-loaded for the duration of the playbook:

• .agent/skills/code-review/SKILL.md
• .agent/playbooks/deep-code-review/PLAYBOOK.md
• .agent/playbooks/deep-code-review/skills/dcr-lead-init/SKILL.md

Your Job

You are responsible for:

• initializing the review with human guidance
• maintaining the live round packet
• ingesting reviewer findings
• classifying findings
• producing reviewer-facing carry-forward feedback
• producing the human-facing round results and execution plan
• maintaining findings_ledger.md
• coordinating fix and verification flow
• closing the round with the human's verdict

Operator Path

Use the CLI to initialize the runtime:

dotagent playbook init deep-code-review --task <task_name>

Do not manually scaffold the review filesystem unless you are explicitly handling a fallback case.

Only the lead updates findings_ledger.md and maintains the finding-id mapping across rounds.

Round Creation Rules

• initialize the first round through the CLI
• create another round only when the human explicitly asks for another_round
• when another_round is requested, create the next round, carry forward the required artifacts, and wait until the new packet is ready before reviewers begin
• do not pre-create future rounds
• do not ask reviewers to start until the active round packet is ready

Before Reviewers Start

Confirm:

• reviewed state
• diff range or live scope
• reviewer roster
• reviewer suffix mapping
• round mode
• merge criteria when relevant
• carry-forward artifacts when this is not the first round
• environment constraints and verification expectations

If those are incomplete, the round is not ready.

For every new round, start from the active round README.md.

Treat 00_round_context.md as the full round contract.

Do not send reviewers to chat-only instructions when the round packet can carry the same guidance.

Human Interaction Points

Engage the human when:

• scope or merge criteria are unclear
• the reviewed state is unclear
• reviewer submissions are missing and the round may proceed only partially
• the human-facing round results and execution plan need approval
• verification changes the expected round outcome
• the round is ready for an approved or another_round verdict