| name | executing-plans |
| description | Executes written implementation plans by loading the plan, reviewing it critically, and running each task in sequence with CI verification. Use when the user has a written implementation plan ready to execute in a dedicated session with human review checkpoints. |
Executing Plans
Overview
Load plan, review critically, execute all tasks, report when complete.
Announce at start: "I'm using the executing-plans skill to implement this plan."
The Process
Step 1: Load and Review Plan
- Find the plan file. Plans live at
docs/<feature>/plans/<feature-name>.md. Glob docs/*/plans/*.md, exclude any file whose name ends in -handoff.md (those are inputs to writing-plans, not plans), and pick the most recent by modification time. If no plan file is found, or if multiple candidates exist and you cannot determine which the user intends, ask: "Please provide the path to the implementation plan file." Do not proceed without a confirmed plan file.
- Read the plan file in full.
- Review critically - identify any questions or concerns about the plan before starting.
- If concerns: Raise them with your human partner before starting.
- If no concerns: Create a task checklist (use
TodoWrite if available; otherwise maintain a checklist in your response) and proceed.
Step 2: Execute Tasks
For each task:
- Mark as in_progress
- Follow each step exactly (plan has bite-sized steps)
- When a plan step references a skill by name, invoke that skill rather than implementing the step manually
- Run verifications as specified
- Mark as completed
Step 3: Complete Development
After all tasks complete and verified, run the full CI verification sequence.
Identify the project's CI commands by checking package.json scripts. Run the equivalent of:
If any step fails, fix the issue before proceeding. Report the results to the
user and summarize: which tasks were completed, which files were changed, and
whether all checks passed.
The author decides when (and whether) to commit — do not initiate git commit, git add, or any git write operation on your own.
When to Stop and Ask for Help
STOP executing immediately when:
- Hit a blocker (missing dependency, test fails, instruction unclear)
- Plan has critical gaps preventing starting
- You don't understand an instruction
- Verification fails repeatedly
Ask for clarification rather than guessing.
When to Revisit Earlier Steps
Return to Review (Step 1) when:
- Partner updates the plan based on your feedback
- Fundamental approach needs rethinking
Don't force through blockers - stop and ask.
Code Conventions
Apply these conventions in all code you write or modify during plan execution. They reflect the project's established patterns and keep the codebase consistent.
1. Use the Issue Factory for Error/Issue Objects
Never construct issue or validation-result objects inline. The project has a dedicated factory (look for IssueFactory, createIssue, or a similar factory pattern in the codebase) so that all issues are defined in one place, are easy to audit, and have consistent shape.
Wrong:
throw new Error('...');
return {
code: 'ARC-SESSION-ALREADY-ACTIVE',
message: `An active session already exists...`,
severity: IssueSeverity.Error,
};
Right:
return IssueFactory.sessionAlreadyActive(cmd.projectId, existing.sessionId, existing.mode);
If the factory doesn't yet have a method for the issue you need, add one to the factory rather than creating the object inline at the call site.
2. Derive Types from Existing Object Literals — Don't Repeat String Unions
When a set of valid string values is already captured in a const object, enum, or similar structure, derive the type from it. Repeating the values as a string literal union (e.g., 'STAGED' | 'UNSTAGED' | 'PARTIAL') creates two sources of truth that can drift.
Wrong:
private computeStatus(rows: Row[]): 'STAGED' | 'UNSTAGED' | 'PARTIAL' { ... }
Right (if a const/enum exists or should exist):
const PendingChangeStatus = { STAGED: 'STAGED', UNSTAGED: 'UNSTAGED', PARTIAL: 'PARTIAL' } as const;
type PendingChangeStatus = typeof PendingChangeStatus[keyof typeof PendingChangeStatus];
private computeStatus(rows: Row[]): PendingChangeStatus { ... }
If no such object exists yet and more than two callers would benefit, introduce one.
3. Prefer TypeORM Query Builder Over Raw SQL
For all database queries and writes, use TypeORM's QueryBuilder, repository methods, or entity manager APIs rather than raw SQL strings — unless profiling shows a measurable performance difference that justifies the trade-off.
Wrong:
await dataSource.query(`SELECT * FROM edit_action WHERE workspace_id = $1`, [id]);
Right:
await dataSource
.getRepository(EditAction)
.createQueryBuilder('ea')
.where('ea.workspaceId = :id', { id })
.getMany();
Raw SQL is acceptable for complex aggregations or queries where the TypeORM abstraction would be significantly harder to read, but note the reason in a comment.
4. No Underscore Prefix for Private Members
TypeScript's private keyword (or # for truly hard-private fields) makes the intent clear. The _ naming convention adds noise without benefit.
Wrong:
private _sessionId: string;
Right:
private sessionId: string;
This applies to all new code and to any existing _-prefixed members you touch while implementing a plan step.
Remember
- Review plan critically first
- Follow plan steps exactly
- Apply code conventions above to all code written or modified
- Don't skip verifications
- Reference skills when plan says to
- Stop when blocked, don't guess
- Never start implementation on main/master branch without explicit user consent
Integration
This skill is the third step in the ARC development workflow:
- brainstorming → Explores requirements and produces a design spec
- writing-plans → Converts the spec into a step-by-step implementation plan
- executing-plans (this skill) → Executes the plan task by task
The plan file is produced by the writing-plans skill and saved to docs/<feature>/plans/.
