om-spec-writing

Write and review feature specifications with skeleton-first drafting and severity-ranked architectural review.

1|Updated Apr 23, 2026
One-click install
npx skills add https://github.com/tkogut/agents-os-core --skill om-spec-writing-tkogut
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: om-spec-writing
Source: https://github.com/tkogut/agents-os-core/tree/main/global_skills/om-spec-writing
Command: npx skills add https://github.com/tkogut/agents-os-core --skill om-spec-writing-tkogut

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Writing feature specifications that are complete, architecturally sound, and ready for implementation is slow and error-prone; this Skill enforces a staff-engineer-grade process with a hard Open Questions gate so specs never ship with unresolved critical unknowns. ## Core Features & Use Cases - Skeleton-first drafting: Start with a minimal TLDR plus 2-3 sections, surface critical unknowns as a numbered Open Questions block, and iterate only after answers arrive. - Structured spec output: Produces dated spec files with Problem Statement, Architecture, Data Model, API Contracts, Edge Cases, Phasing, and a step-level Implementation Plan that feeds directly into automated PR creation. - Severity-ranked architectural review: Reviews existing specs against repository agent instructions, canonical primitives, backward-compatibility contracts, and reversibility, reporting Critical/High/Medium/Low findings with evidence and corrections. - Autonomous mode: An --autonomous flag resolves gate questions with the most reversible defaults and marks risky ones as needing human confirmation, enabling unattended pipeline runs. - Use Case: A developer starting a new billing feature invokes the skill, answers three gate questions about scope and data model, and receives a phased spec whose implementation plan is handed phase-by-phase to an automated PR-creation workflow. ## Quick Start Ask the agent to write a spec for your new feature using the om-spec-writing skill and answer the Open Questions it presents.

Frequently Asked Questions about om-spec-writing

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

FAQPage Schema
How do I write a feature specification with an AI agent?▼

Invoke the skill with your feature idea; it loads repository architecture rules, drafts a minimal skeleton spec, and presents numbered Open Questions for critical unknowns. After you answer, it iterates into a full spec with architecture, data model, and a phased implementation plan.

How do I review a spec for architectural problems?▼

Point the skill at an existing spec file and it applies a staff-engineer review checklist covering scope cohesion, canonical primitives, backward compatibility, reversibility, and failure scenarios. Findings are ranked Critical through Low with evidence and recommended corrections.

Can spec writing run autonomously without user interaction?▼

Yes, the --autonomous flag resolves Open Questions itself using the most reversible, lowest-blast-radius defaults and records them in a Resolved assumptions section. Risky defaults are marked as needing human confirmation and reported to the calling workflow.

Where are generated spec files stored?▼

Specs are written as dated Markdown files named YYYY-MM-DD-kebab-case-title.md inside the specs directory resolved from .ai/agentic.config.json, defaulting to .ai/specs. Without config it falls back to existing design-doc directories like docs/specs or rfcs.

What are the limitations of automated spec review?▼

The review examines the document against repository rules and cited evidence but does not test the implementation. It explicitly reports review limits, unchecked areas, and distinguishes verified defects from product direction decisions that need a human call.