GoalPro v0.1.3

Clarifies that GoalPro's deliverable is always two copyable instruction prompts: Goal Prompt and Loop Prompt.

Changed

• Reframe Workflow lens as a judgment layer used while generating the two prompts, not as a third output, workflow prompt, executor, scheduler, or flowchart deliverable.
• Make repeated-work guidance write Trigger, Checkpoint, Brief, and source-of-truth details into Goal Prompt / Loop Prompt only when they help the two-prompt contract.
• Update the website and Xiaohongshu examples so they explicitly start from the two-prompt output shape.
• Mirror the clarification across .agents and .claude.

Verification

• git diff --check passed with only LF-to-CRLF warnings.
• .agents and .claude GoalPro skill mirrors have matching SHA256 hashes for SKILL.md, references/source-rules.md, and references/examples.md.
• Key dual-prompt boundary phrases were checked with rg, including Goal Prompt + Loop Prompt, 第三个 Workflow Prompt, and 两段指令提示词.
• graphify update . completed after the wording correction.

GoalPro v0.1.2

Adds a workflow lens to GoalPro so repeated automation-style requests become safer, clearer Goal + Loop contracts without turning every task into a workflow.

Changed

• Add Workflow lens guidance for recurring work such as daily, weekly, automatic, publishing, operations, monitoring, queues, review, and sync tasks.
• Teach GoalPro to add Trigger, Checkpoint, Brief, and source-of-truth guidance only when a request is genuinely repeatable.
• Strengthen the one-question rule: route-changing ambiguity should ask one blocking question with a recommended answer, while context-answerable questions should be resolved by reading context.
• Update the Xiaohongshu publisher example to model a safe publishing workflow with late human confirmation, decision briefs, and clear non-automation boundaries.
• Clarify that one-off website work should not be forced into workflow form unless the user asks for ongoing feedback, updates, operations, or measurement.
• Mirror all Skill, source-rules, and examples changes across .agents and .claude.

Verification

• git diff --check passed with only LF-to-CRLF warnings.
• .agents and .claude GoalPro skill mirrors have matching SHA256 hashes for SKILL.md, references/source-rules.md, and references/examples.md.
• Key workflow-lens fields were checked with rg, including Workflow lens, Trigger, Checkpoint, Brief, 推荐答案, and 小红书自动发布器.
• graphify update . completed after the workflow-lens update.

GoalPro v0.1.1

Improves GoalPro's Loop Prompt contract and adds concrete, user-judgeable examples for automation-heavy fuzzy product requests.

Changed

• Front-load 时间参数 in Loop Prompt guidance while keeping automation setup as a separately authorized step.
• Strengthen completion reporting so final reports include a judgeable sample or case snippet, not just validation claims.
• Refine Loop rules around evidence review, intent amplification, automation readiness, and stop or escalation boundaries.
• Add a full Xiaohongshu publisher example that requires deep research before implementation and distinguishes publishing assistance from unsafe or unverified automation.
• Mirror all GoalPro Skill, source-rules, and examples changes across .agents and .claude.

Verification

• git diff --check passed with only LF-to-CRLF warnings.
• .agents and .claude GoalPro skill mirrors have matching SHA256 hashes for SKILL.md, references/source-rules.md, and references/examples.md.
• Key fields and examples were checked with rg, including 时间参数, 验收样例, and 我需要小红书自动发布器.
• graphify update . completed after the example update.

GoalPro v0.1.0

Changelog

v0.1.0 - 2026-06-26

GoalPro's first GitHub release formalizes the upgrade from a single Goal Contract generator into a dual-prompt Skill for execution and continued improvement.

Changed

• Reframed GoalPro around Goal Prompt + Loop Prompt across the README and both runtime skill copies.
• Kept the prompt-only boundary explicit: GoalPro generates copyable prompts, but does not execute the goal or run the loop without separate user authorization.
• Added Loop fields for evidence review, gap diagnosis, verification delta, guardrails, Done / Continue / Pause decisions, and Next LOOP packet handoff.
• Expanded examples for login, strategic research, file + chat output, fuzzy website requests, TypeScript migration, and Claude Code task prompts.
• Updated source rules to connect iterative improvement with evidence, loop state, no-progress thresholds, and stop/escalation conditions.
• Ignored .claude/project-task-state.json so local Claude task state stays out of Skill releases.

Verification

• .agents and .claude GoalPro skill mirrors have matching SHA256 hashes for SKILL.md, references/source-rules.md, and references/examples.md.
• Static diff inspection confirms the release scope is limited to GoalPro docs, skill rules, examples, and release notes.
• graphify query was used before release-oriented source inspection, following the repo's graph-first navigation rule.