execute-plan

Overview

This skill dispatches the dev-workflow:execute-plan agent (sonnet) to execute a verified plan. The agent follows plan tasks precisely without making judgment calls. The plan's final verification task runs the full build/test suite. Failure fixes happen after the agent returns, in main context.

Process

Step 1: Pre-checks

1. Read the plan file
2. Verification pre-check: Look for a ## Verification section with Verdict: Approved in the plan file
   • If found: verification is done, continue
   • If not found: invoke dev-workflow:verify-plan before proceeding. If verify-plan returns "must-revise", apply revisions and re-verify before continuing
3. Decision Points: If the plan file contains a ## Decisions section with unresolved decisions (no **Chosen:** line), present them before dispatching:
   • For each blocking decision: present to user via AskUserQuestion with options from the decision point
   • For recommended decisions: present as a group via a single AskUserQuestion. Critical: all DP content must be inside the question field — text printed before AskUserQuestion gets visually covered by the question widget. Read each recommended DP's full block (heading + Context + Options + Recommendation) from the plan file and concatenate them verbatim in the question field, separated by \n---\n. End with: \n\n全部接受推荐，还是逐个审查？
   • If the user does NOT choose to accept all: present each DP individually via separate AskUserQuestion calls. Do not assume any DP is accepted until the user explicitly confirms it
   • Record user choices: edit the plan file, replace the **Recommendation:** or **Recommendation (unverified):** line with **Chosen:** {user's choice}

Step 2: Dispatch Agent

Use the Agent tool to dispatch the dev-workflow:execute-plan agent:

Execute the implementation plan.

Plan file: {plan file path}
Project root: {project root}

Step 3: Process Results

When the agent returns:

1. Read the report file at docs/06-plans/execution-report.md
   • If the agent's return message contains Report written to:, use that path
   • If the file exists but **Status:** is in-progress, the agent was truncated — the file still contains all completed task results up to the truncation point
   • If the file does not exist: fall back to parsing the agent's return message
2. Present summary to the user: completed/blocked/failed counts
3. If blocked or failed tasks exist: list them with reasons

Standalone mode (not within run-phase):

• If failures: fix in main context (opus)
• Suggest dev-workflow:implementation-reviewer for plan-vs-code audit
• Suggest dev-workflow:finish-branch for branch integration

State Integration

When running within a phase orchestrated by run-phase:

If .claude/dev-workflow-state.yml exists and phase_step is execute:

• After agent returns, do NOT update phase_step (orchestrator owns state transitions)
• Output: "Execution complete. Returning to run-phase."

Completion Criteria

• All plan tasks attempted by the agent
• Execution report reviewed and presented to user
• When in run-phase context: handoff signal output
• When standalone: wrap-up suggestions presented