Audit and rewrite technical specifications for clarity and testability.
npx skills add https://github.com/codacy/codacy-specs --skill spec-improve
Please help me install this Agent Skill. Skill: spec-improve Source: https://github.com/codacy/codacy-specs/tree/main/skills/spec-improve Command: npx skills add https://github.com/codacy/codacy-specs --skill spec-improve
This skill solves the problem of vague, ambiguous, or incomplete specifications that cause AI agents to make incorrect assumptions, fail to implement requirements, or produce unverifiable code.
Use the spec-improve skill to audit the specification in the current file and rewrite it to be ready for implementation.
High-intent search queries and answers about installing and using this skill.
To audit Jira or Linear tickets for AI-agent readiness, this skill detects spec types and scores them on a 0–100 rubric across critical dimensions like acceptance criteria, error handling, and testability. It requires integration with issue tracking MCP servers to perform full audits.
The best way to rewrite vague technical specifications is to use a guided clarification process that asks focused questions to fill gaps, then produces a fully structured, AI-agent-ready specification using proven templates.
Yes, spec-improve works with local file sources and GitHub. It audits and rewrites technical specifications across Jira, Linear, GitHub, and local files by applying quality rubrics to ensure requirements are clear and unambiguous.
Specification gaps are identified through a scoring rubric, then clarified via guided questions asked one at a time. This focused clarification loop fills critical gaps in requirements before rewriting the spec.
Yes, you need issue tracking MCP servers and local file access to perform full audits, clarification loops, and write-back updates across platforms like Jira and Linear.
You should audit specs before starting work on a Jira ticket or local spec file to ensure the requirements are clear, testable, and unambiguous, preventing AI agents from making incorrect assumptions or failing to implement requirements.
jpoutrinConfigure MCP servers for Claude Code with HTTP, SSE, and stdio transports.
shynlee04Run AI agents in a loop to implement and verify user stories.
foomakersCreate or update issues in GitHub Projects, Jira, or Linear using templates.
silaslopes100Decompose product requirements documents into structured Spec-Driven Development components.
