_shared

Stores shared SDD reference documents for persistence, status, and review contracts.

Updated Aug 25, 2026
One-click install
npx skills add https://github.com/CarlosWilliamsR/SketchOS --skill shared-carloswilliamsr
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: _shared
Source: https://github.com/CarlosWilliamsR/SketchOS/tree/main/.config/opencode/skills/_shared
Command: npx skills add https://github.com/CarlosWilliamsR/SketchOS --skill shared-carloswilliamsr

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? SDD phase skills need consistent rules for artifact persistence, status reporting, skill resolution, and review orchestration, and duplicating those rules in every skill causes drift and contradictions. ## Core Features & Use Cases - Shared Reference Documents: Centralizes conventions for Engram memory persistence, OpenSpec file layouts, and the persistence contract covering engram, openspec, hybrid, and none modes. - Common Phase Protocol: Defines the standard SDD phase boilerplate including skill loading, artifact retrieval, return envelopes, and review workload guards. - Status and Review Contracts: Documents the SDD status schema, native dispatcher integration, and the bounded review orchestration lifecycle. - Use Case: When building or maintaining an SDD phase skill such as sdd-propose or sdd-verify, read these references to align persistence calls, status output, and sub-agent delegation with the shared protocol. ## Quick Start Read the relevant shared reference file, such as persistence-contract.md or sdd-phase-common.md, before implementing or invoking any SDD phase skill.

Frequently Asked Questions about _shared

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

FAQPage Schema
How do SDD skills persist artifacts across sessions?

SDD skills persist artifacts using the persistence contract with four modes: engram, openspec, hybrid, or none. Engram mode saves to memory with deterministic topic keys, openspec writes files under openspec/changes, and hybrid writes to both.

What is the difference between engram and openspec artifact stores?

Engram provides cross-session recovery and compaction survival but overwrites on upsert with no history. OpenSpec creates git-committable files with full audit trails but no cross-session memory. Hybrid combines both at higher token cost.

Can the _shared skill be invoked directly?

No, _shared is a support package only and is explicitly marked as not invokable with disable-model-invocation set to true. It stores reference documents consumed by real SDD skills such as sdd-propose and sdd-verify.

Why does mem_search require a follow-up retrieval call?

mem_search returns only 300-character previews, not full content. Every artifact retrieval requires a second mem_get_observation call with the saved observation ID to obtain complete content, otherwise downstream phases produce wrong output.

How do sub-agents receive skill context in SDD workflows?

The orchestrator resolves skill paths from the skill registry and injects them as a Skills to load before work block in the launch prompt. Sub-agents read those exact SKILL.md files and report their resolution method as paths-injected, fallback-registry, fallback-path, or none.