executing-plans

Executes written implementation plans task-by-task with review checkpoints and a durable progress ledger.

Updated May 21, 2026
One-click install
npx skills add https://github.com/CagesThrottleUs/private-ai-harness --skill executing-plans-cagesthrottleus
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: executing-plans
Source: https://github.com/CagesThrottleUs/private-ai-harness/tree/main/skills/executing-plans
Command: npx skills add https://github.com/CagesThrottleUs/private-ai-harness --skill executing-plans-cagesthrottleus

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Written implementation plans often get executed inconsistently across sessions: tasks get skipped, progress is lost after context compaction, and verification steps are forgotten. This Skill enforces a disciplined execution loop that loads a plan, reviews it critically, executes every task with verifications, and records progress in a durable git-based ledger. ## Core Features & Use Cases - Durable progress tracking: Writes completed tasks to a ledger under the git directory so execution can resume after session compaction without re-running finished work. - Pre-commit quality gates: Runs design-principles review, karpathy anti-pattern checks, code-documentation requirements, and a linter gate before every task commit. - Conditional skill invocation: Automatically invokes integration-testing for components with external dependencies and observability-standards for new API endpoints. - Use Case: You have a multi-task implementation plan written in a previous session. Invoke this Skill in a fresh session; it resumes from the first incomplete task, executes each step, and finishes with the branch-completion workflow. ## Quick Start Use the executing-plans skill to implement the plan in docs/plans/feature-auth.md, resuming from the progress ledger if one exists.

Frequently Asked Questions about executing-plans

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I execute an implementation plan across multiple sessions?

Invoke the executing-plans skill with your plan file. It checks a progress ledger stored under the git directory, skips tasks already marked complete, and resumes at the first incomplete task so work survives session compaction.

How to resume plan execution after context compaction?

The skill reads a ledger at the git path executing-plans/progress.md before doing anything else. Tasks listed as complete are not re-executed, and execution resumes at the first incomplete entry without relying on conversation memory.

What quality checks run before each task commit?

Before committing, the skill runs the design-principles review checklist, applies the karpathy lens for surgical spec-traceable changes, requires full code documentation with spec_id and req_id, and enforces a zero-issue linter gate with format, lint, and type checks.

When should I stop executing a plan and ask for help?

Stop immediately when you hit a blocker such as a missing dependency, failing test, unclear instruction, or critical plan gap. Ask for clarification rather than guessing, and never force through repeated verification failures.

Does executing-plans work without subagent support?

Yes, but the skill notes that quality is significantly higher on platforms with subagent support such as Claude Code or Codex. If subagents are available, it recommends using subagent-driven-development instead.