skills.rest
mnfst-ai
mnfst-aiOFFICIAL
Explore all skills by this author →
WebsiteGitHub

sm:stage:decision-capture

Generate a prioritized decisions-made manifest classifying choices as Specified, Invented, or Adapted.

5|2|Updated Mar 8, 2026
Search Tags:#architecture#audit#pipeline#manifest#code-review#git-diff#decision-capture
One-click install
npx skills add https://github.com/mnfst-ai/Stage_Manager_Skills --skill sm-stage-decision-capture
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: sm:stage:decision-capture
Source: https://github.com/mnfst-ai/Stage_Manager_Skills/tree/main/plugins/stage-manager/skills/stage-decision-capture
Command: npx skills add https://github.com/mnfst-ai/Stage_Manager_Skills --skill sm-stage-decision-capture

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill reads a project's plan and the complete build output to produce a decisions-made manifest that names every implementation choice the coding tool made that was not explicitly specified, turning invisible inventions into actionable items for the builder.

Core Features & Use Cases

  • Full-build analysis: Compare the original plan to the full git diff, PR, or set of files produced across sessions to catalog implementation choices.
  • Classification & prioritization: Label each item as Specified, Invented, or Adapted and surface P1/P2/P3 priority decisions with consequences and recommended actions.
  • Pipeline-ready output: Emit a structured YAML feed-forward block for downstream /ce:review and /ce:compound steps, and provide a clear close/next-action flow.
  • Use Cases: Run after a multi-session build to discover what the agent added, before a formal code review to provide context, or to audit architectural drift across a project.

Quick Start

Compare the plan file and the build diff and generate a prioritized decisions-made manifest that lists specified, invented, and adapted items.

Frequently Asked Questions about sm:stage:decision-capture

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

FAQPage Schema
How do I audit architectural drift after a multi-session build?▼

To track unspecified implementation choices, compare your original plan file against the full build output or git diff to generate a decisions-made manifest. This classifies items as Specified, Invented, or Adapted for review.

What is a decisions-made manifest in code review?▼

A decisions-made manifest identifies every implementation choice a coding tool made that was not explicitly specified. It turns invisible inventions into a prioritized, structured YAML block for downstream review.

How do I generate a structured YAML feed-forward block for code review?▼

Generate a structured YAML feed-forward block by applying the Skill to completed builds, pull requests, or git diffs. It compares the original plan against produced files to classify decisions for downstream review.

Can I use git diff to find invented code choices not in the original plan?▼

Yes, you can use a git diff to find invented choices by comparing it against the original plan file. The manifest labels each implementation choice as Specified, Invented, or Adapted to surface deviations.

Do I need an original plan file to audit implementation choices?▼

Yes, you need the original plan file and the full build output to audit implementation choices. The Skill requires both inputs to compare the specified plan against produced files and classify decisions.

How do I classify decisions as Specified, Invented, or Adapted for a pipeline?▼

Classify decisions as Specified, Invented, or Adapted by comparing the plan file against the full build output. The manifest assigns P1/P2/P3 priorities with consequences and recommended actions for pipeline-ready output.

Related Skills

Explore More Software Engineering Skills→
LorenFrankLabLorenFrankLab
3

safe-refactoring

Refactor code structure while preserving behavior via baseline tests and numerical equivalence.

Official
Intermediate
rohunjrohunj
232

compound-engineering

Automate the Plan, Work, Review, and Compound loop for development projects.

Community
Advanced
TonyQ-AITonyQ-AI
9

requesting-code-review

Deploy a sub-agent to review git diffs against defined requirements.

Community
Advanced
ClearMeasureLabsClearMeasureLabs
3

stylecop

Analyze C# source code for style violations using a clean-architecture rule profile.

Official
Intermediate