speckit-converge

Assess codebase against spec, plan, and tasks, then append remaining work to tasks.md.

Updated Jun 27, 2026
One-click install
npx skills add https://github.com/marionoratoRN/custom-speckit-flow --skill speckit-converge-marionoratorn
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: speckit-converge
Source: https://github.com/marionoratoRN/custom-speckit-flow/tree/main/.claude/skills/speckit-converge
Command: npx skills add https://github.com/marionoratoRN/custom-speckit-flow --skill speckit-converge-marionoratorn

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? After running an implementation pass, teams often cannot tell which requirements, acceptance criteria, or plan decisions remain unbuilt. This Skill closes that gap by comparing the current codebase against the feature's spec.md, plan.md, and tasks.md and appending any unfinished work as new traceable tasks. ## Core Features & Use Cases - Gap Assessment: Reads spec.md, plan.md, and tasks.md as the source of intent and classifies findings as missing, partial, contradicts, or unrequested work. - Severity Grading: Ranks findings CRITICAL through LOW, treating constitution MUST violations as the highest severity. - Append-Only Task Generation: Adds a new numbered Convergence phase to tasks.md with zero-padded task IDs, never rewriting or renumbering existing tasks. - Use Case: After /speckit-implement finishes, run this Skill to discover that FR-008's append-only guard was never built; it appends a HIGH-severity task so the next implement pass completes it. ## Quick Start Ask the assistant to run speckit-converge on the current feature to find any unimplemented requirements and append them to tasks.md.

Frequently Asked Questions about speckit-converge

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

FAQPage Schema
How do I find unimplemented requirements after running speckit-implement?▼

Run the speckit-converge command after /speckit-implement completes. It reads spec.md, plan.md, and tasks.md, inspects the code in scope, and appends each remaining gap as a new task under a Convergence phase in tasks.md.

What does speckit-converge write to tasks.md?▼

It appends a single new section titled '## Phase N: Convergence' containing one checklist item per finding, with zero-padded IDs continuing from the highest existing task number. It never modifies, reorders, or deletes existing tasks, and leaves the file untouched when nothing remains.

When should I run speckit-converge in the spec-kit workflow?▼

Run it only after /speckit-tasks has produced a complete tasks.md and /speckit-implement has executed against it. It requires spec.md, plan.md, and tasks.md to exist and stops with a message naming the prerequisite command if any are missing.

Does speckit-converge modify spec.md, plan.md, or application code?▼

No. Its only write operation is appending a Convergence phase to tasks.md. It never edits spec.md or plan.md and never creates or deletes application code; completing the appended tasks is the job of /speckit-implement.

How does speckit-converge handle constitution violations?▼

Code violating a MUST principle in .specify/memory/constitution.md is classified as CRITICAL severity, the highest level. Constitution-violation tasks are emitted first in the appended phase, and if the constitution is an unfilled template the check is skipped gracefully.

What happens when the codebase already satisfies the spec?▼

The command reports a converged outcome, leaves tasks.md byte-for-byte unchanged with no empty phase header, and recommends proceeding to review or opening a pull request instead of running another implement pass.