Can I customize the staleness threshold?

Yes, the skill reads from 'ops/config.yaml' to determine the specific day-threshold for when a decision is considered stale and requires a new review.

What is the 'Claim Test' for titles?

The Claim Test ensures a title is a full argument (e.g., 'Database X requires Y architecture') rather than a vague topic (e.g., 'Database notes'), making the decision-making process transparent.

Does this skill automatically change my decision notes?

No. While it can auto-apply mechanical updates like 'last_reviewed' dates, it only proposes changes for titles and bodies to ensure human oversight of reasoning and intent.

How does the skill verify reasoning quality?

It analyzes the text for connective logic words like 'because', 'therefore', and 'however' to ensure the entry explains the path to a conclusion rather than just asserting it.

PM Decision Reviewer

Name: PM Decision Reviewer
Author: LopeWale

byLopeWale

•

生产力与工作流

Audits and enforces quality standards for architectural and technical decision registers to ensure reliable project documentation.

PM Decision Reviewer is a specialized tool for Claude Code designed to maintain the integrity of project decision logs and ADRs (Architectural Decision Records) through rigorous structural and reasoning audits. Unlike basic linters, it evaluates the logic of your documentation by performing 'claim tests' on titles and 'reasoning tests' on content, ensuring every entry provides clear, verifiable justification. This skill is essential for complex projects where stale or poorly reasoned decisions can lead to technical debt or misinformed AI inferences, helping teams maintain a trustworthy and navigable history of project evolution.

主要功能

01Reasoning audit to verify the presence of logical connectives and evidence

02Staleness tracking to identify outdated documentation and status mismatches

03Schema compliance validation for metadata fields and decision types

041 GitHub stars

05Connection completeness checking to link decisions to specific sprint records

06Claim Test title evaluation to ensure titles state a clear, actionable position

使用场景

01Identifying 'orphan' decisions that haven't been reviewed or linked to a project register

02Standardizing decision-making documentation across a distributed engineering team

03Performing a quality audit on an ADR repository before a major system migration

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add lopewale/dev_ignition pm-review

For use in Claude.ai and ChatGPT