skills.rest
lokafinnsw
lokafinnswCOMMUNITY
Explore all skills by this author →
GitHub

using-goal-mode

Guide controller agent operations in Claude Code goal-mode lifecycle states.

3|1|Updated May 9, 2026
Search Tags:#claude-code#acceptance-criteria#evidence-tracking#review-gates#prompt-guardrails#goal-mode#lifecycle-control
One-click install
npx skills add https://github.com/lokafinnsw/claude-code-goal-mode --skill using-goal-mode
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: using-goal-mode
Source: https://github.com/lokafinnsw/claude-code-goal-mode/tree/main/skills/using-goal-mode
Command: npx skills add https://github.com/lokafinnsw/claude-code-goal-mode --skill using-goal-mode

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Prevents mistakes that break or confuse a Claude Code goal-mode run, including incorrect tag emission, fabricated reviewer verdicts, and recovery failures across lifecycle transitions and multi-session setups.

Core Features & Use Cases

  • Guides controller behavior in goal-mode: explains how to interact with the plan-tree engine during active pursuing, review-pending, blocked, and awaiting-manual-approval states.
  • Enforces correct evidence and achievement flow: shows how to map acceptance criteria to concrete evidence and when it is valid to claim achievement.
  • Maintains reviewer independence and escape-hatch recovery: instructs how to dispatch reviewer subagents and how to handle missing reviewer agents using the REVISE-unavailable pattern that routes to manual approval.

Quick Start

Use the using-goal-mode skill at the start of a goal-mode run to follow the correct interaction protocol for cursor inspection, evidence capture, reviewer dispatch, and safe lifecycle transitions.

Frequently Asked Questions about using-goal-mode

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

FAQPage Schema
How do I handle Claude Code goal-mode lifecycle transitions without breaking the state machine?▼

Handling Claude Code goal-mode lifecycle transitions safely requires maintaining strict tag discipline or using explicit slash commands, ensuring correct Stop-hook versus controller control modes across pursuing, review-pending, blocked, and awaiting-manual-approval states to prevent state machine corruption.

How does evidence tracking work when claiming achievement in a goal-mode run?▼

Evidence tracking in a goal-mode run works by mapping acceptance criteria to concrete evidence captured during plan-tree execution, allowing achievement claiming only when that specific evidence is verified and properly recorded during the active pursuing state.

What happens if a reviewer subagent is missing during a Claude Code goal-mode review?▼

When a reviewer subagent is missing during a Claude Code goal-mode review, you should apply the REVISE-unavailable pattern, which routes the process to manual approval recovery while maintaining reviewer independence through an approved escape-hatch protocol.

When should I use Stop-hook control mode versus controller control mode in Claude Code goal-mode?▼

Stop-hook control mode versus controller control mode in Claude Code goal-mode is determined by the current command phase and lifecycle state, requiring strict adherence to the interaction protocol to avoid fabricated reviewer verdicts and multi-session recovery failures.

Why do my goal-mode runs fail with fabricated reviewer verdicts and how can I prevent it?▼

Goal-mode runs fail with fabricated reviewer verdicts when reviewer independence is not properly enforced, which you can prevent by dispatching reviewer subagents correctly and applying the REVISE-unavailable escape-hatch protocol for manual approval recovery.

Related Skills

Explore More Software Engineering Skills→
adolfoaranaes12adolfoaranaes12
2

trace-requirements

Generate bidirectional traceability matrices mapping acceptance criteria to implementation evidence and tests.

Community
Advanced
Liuchun-ossLiuchun-oss
15

subagent-driven-development

Dispatch independent subagents for isolated task execution and two-stage review.

Community
Advanced
ylubiylubi
12

ralph-planner

Orchestrate Planning, Implementation, and Testing phases across three rounds.

Community
Advanced
JaCha00JaCha00

ralph

Run complex tasks until completion with architect verification and state tracking.

Community
Advanced