B2C Operations Runbooks
Agent skills for administrators, operators, and developers investigating B2C
Commerce health and incidents. Available as the b2c-ops plugin and bundled in
the B2C MCP. No demo workspace or application source is required.
| Skill | Outcome |
|---|---|
b2c-production-triage | Scope an incident, correlate evidence, and prepare a handoff. |
b2c-job-health | Review scheduled work and investigate failed or incomplete updates. |
b2c-checkout-triage | Investigate broader checkout/payment symptoms and select relevant evidence. |
b2c-order-failure-triage | Enumerate FAILED orders, quantify impact, and correlate failed inputs with logs and jobs. |
b2c-integration-triage | Investigate failing, slow or silent calls to external services using third-party metrics, logs and service config. |
b2c-performance-triage | Localize slow or erroring APIs/pages with Metrics API signals (family, controller, hook, host, cache). |
b2c-edge-traffic-triage | Classify eCDN traffic (bots, abuse, edge errors), check edge protections, propose rules. |
See the Operations guide for setup and example requests. Direct plugin installation adds guidance; it does not grant access or install the MCP/CLI. MCP users need no separate install.
Installation
# Claude Code
claude plugin marketplace add SalesforceCommerceCloud/b2c-developer-tooling
claude plugin install b2c-ops@b2c-developer-tooling
# GitHub Copilot CLI
copilot plugin marketplace add SalesforceCommerceCloud/b2c-developer-tooling
copilot plugin install b2c-ops@b2c-developer-tooling
Codex: open the repo as a workspace, restart Codex, then install from the B2C Developer Tooling marketplace in the plugin directory.
For file-copy install to any supported IDE, use b2c setup skills b2c-ops. See the install guide for details.
Adding a runbook
Follow the repository's MCP development skill and workflow authoring reference.
Create skills/<task>/SKILL.md with a descriptive trigger, observable outcome,
MCP/CLI choices, bounded investigation, intervention checkpoint, and recovery or
handoff. Put conditional detail in references/; keep each entry independently
discoverable. Use installed skill names for native assistants and exact skill
URIs for MCP. Verify capabilities against code and current contracts.
Each addition needs an agent-plugin changeset and acceptance scenarios that include missing access and partial evidence. Update the Operations guide when it adds a capability worth highlighting; do not mirror the skill inventory there. Record tooling gaps separately; do not make a new tool just to mirror a runbook. Never include demo answers, tenant defaults, or customer data.