mcsl-domain-core - SKILL.md Agent Skill

name: mcsl-domain-core description: Use when working inside the MCSLDomainExpert project and the user asks anything about PluginHive MCSL across Shopify, WooCommerce, BigCommerce, Magento, and PrestaShop, MCSL app QA domain, app flows, MCSL carrier/API behavior, project architecture, local RAG/code/wiki knowledge, or wants research-backed answers that may require browsing beyond the current knowledge base. This is the shared domain/research core for AC, TC, AI QA, automation, handoff, and support tasks.

MCSL Domain Core

Use this skill as the shared knowledge and research layer for the MCSLDomainExpert project.

It should make Codex/Claude behave like the project Domain Expert:

know the MCSL platform architecture, with Shopify as the current live AI QA/automation surface
understand dashboard pipeline stages
use local project knowledge before guessing
browse official/current sources when local knowledge is missing or stale
cite where facts came from
feed research-backed conclusions into US/AC, TC, AI QA, automation, and handoff work

First Reads

Always start with:

AGENTS.md
CLAUDE.md if extra session context is needed
skills/mcsl-domain-core/references/research_workflow.md

Then read only the project files directly relevant to the task.

What This Skill Covers

Use for questions or tasks about:

PluginHive MCSL behavior across Shopify, WooCommerce, BigCommerce, Magento, and PrestaShop
PluginHive app UI flows
Shopify admin vs app iframe navigation
platform-specific admin/storefront navigation for docs and manual QA when a card names WooCommerce, BigCommerce, Magento, or PrestaShop
ORDERS / Order Summary label generation
automation-rule label generation
app iframe options and settings
request/response logs
Print Documents vs Download Documents
MCSL carrier request/response fields and constraints
packaging, pickup, return label, bulk label, order grid, products, settings
generated AC/TC correctness
AI QA evidence strategy
automation/POM/spec patterns
support/business handoff facts
research that local RAG may not have yet

Research Order

Use this order:

AGENTS.md and local skills
local project files
automation repo files under MCSL_AUTOMATION_REPO_PATH
local wiki / docs / Chroma-backed context if available
official/current web sources when needed

Browse the web when:

the user asks to research, browse, verify, or use latest/current information
local knowledge does not answer the question
MCSL/Shopify/PluginHive rules may have changed
public docs/API behavior is needed for AC/TC correctness
a linked PR, docs page, Zendesk, changelog, or issue is referenced and its content is not already provided

For web research, prefer:

official PluginHive MCSL knowledge-base pages
official PluginHive docs/help pages
official Shopify docs
project-linked PRs/issues/docs if accessible

Avoid relying on random blogs unless no official source exists, and clearly mark any inference.

Answer Style

For Q&A:

answer directly
mention local/project source and web source when used
separate known fact from inference
include exact file references for local code facts
include links for web sources

For generation tasks:

summarize the research that matters
use the research to improve the output
do not dump long source notes unless the user asks

Relationship To Other Skills

Other MCSL skills should use this skill's research posture:

mcsl-trello-operator: fetch the real card/list/comments/members first when the user gives Trello references.
mcsl-ac-writer-reviewer: research first, then generate/review US + AC, then Trello comment only.
mcsl-dashboard-tc-publisher: generate dashboard TCs, compact Trello comment, and positive-only CSV rows for the Ai tab.
mcsl-ai-qa-testcase-prep: create detailed AI QA executable TCs when browser verification needs richer steps.
mcsl-ai-qa-browser: verify reviewed TCs in Chrome with evidence, cleanup, and locator trace handoff.
mcsl-automation-writer: use reviewed TCs plus AI QA evidence/locator traces to write Playwright automation in MCSL_AUTOMATION_REPO_PATH.
mcsl-bug: format QA-found bugs, check Backlog duplicates, and create Trello Backlog cards only when asked.
mcsl-signoff-message: fetch release/list cards, ask for Backlog links, prepare the QA sign-off message, and send to Slack only after QA confirms channel/message.
mcsl-handoff-docs: generate Support Guide and/or Business Brief PDFs from approved cards.
mcsl-slack-operator: search users/channels, read messages, reply in threads, and send DMs/channel posts only when asked.
mcsl-rag-sync: pull latest and safely sync/reindex StorePep server/client, automation, and wiki knowledge.
mcsl-knowledge-maintainer: after the card cycle, update approved-card RAG, QA feedback, and outdated durable rules.

Normal card-cycle order:

Trello card/list/comments
  -> domain research
  -> US + AC comment
  -> dashboard TCs + Trello/CSV publish package
  -> AI QA browser verification + locator trace
  -> bug follow-up if needed
  -> automation writer
  -> sign-off message
  -> handoff docs
  -> RAG sync if source repos/docs changed
  -> knowledge maintainer

Use Trello/Slack operator skills for actual external reads/writes. Generation skills should prepare content; operator skills should perform Trello/Slack actions when the user clearly asked for those actions.

Do Not

Do not invent MCSL limits or API rules.
Do not assume local RAG is complete.
Do not use stale memory for current MCSL/Shopify/PluginHive rules if browsing is available and relevant.
Do not update Trello, Slack, Sheets, or repo files unless the user asks for that action.
Do not browse for secrets or private data.