Salesforce — SfSkills library router
You are working with a Salesforce library that is far too large to
load. This file tells you how to reach the one page you need.
Generated by scripts/build_plugin.py. Do not hand-edit.
What is in the library
- 1,040 skill packages under
${CLAUDE_PLUGIN_ROOT}/skills/<domain>/<slug>/.
Each is a SKILL.md plus references/examples.md, gotchas.md,
well-architected.md and llm-anti-patterns.md.
- 70 run-time agents under
${CLAUDE_PLUGIN_ROOT}/agents/<id>/AGENT.md,
exposed as subagents (see the roster at the bottom of this file).
- 92 slash commands under
${CLAUDE_PLUGIN_ROOT}/commands/.
- Decision trees under
${CLAUDE_PLUGIN_ROOT}/standards/decision-trees/,
which route between technologies before a skill is opened.
- Canonical templates under
${CLAUDE_PLUGIN_ROOT}/templates/.
Skill counts by domain:
| Domain | Skills | Router |
|---|
admin | 261 | salesforce-admin |
agentforce | 53 | salesforce-agentforce |
apex | 159 | salesforce-apex |
architect | 106 | salesforce-architect |
data | 101 | salesforce-data |
devops | 70 | salesforce-devops |
flow | 63 | salesforce-flow |
integration | 61 | salesforce-integration |
lwc | 83 | salesforce-lwc |
omnistudio | 34 | salesforce-omnistudio |
security | 49 | salesforce-security |
| total | 1,040 | |
How to find the right skill
Three mechanisms, listed in order of reliability on a fresh install.
Use the first one that is available; do not stop at a guess.
1. The shipped rosters (always work, no setup).
Hand off to the domain router below; each one ships a complete
references/skill-index.md roster of its packages, generated from
registry/skills.json. To read one directly, open
${CLAUDE_PLUGIN_ROOT}/.claude/skills/salesforce-<domain>/references/skill-index.md.
2. The MCP server (fast, needs the sfskills-mcp server connected).
Call the search_skill tool with the user's phrasing. It returns
ranked skill ids. get_skill then returns the package contents.
3. The search CLI (fast, needs a locally built index).
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/search_knowledge.py" "<the user's question>" --json
This needs vector_index/, which is not shipped — it is gitignored
and must be built once per clone:
cd "${CLAUDE_PLUGIN_ROOT}" && python3 -m pip install -r requirements.txt && python3 scripts/build_index.py
If the command errors or reports Coverage: NONE, fall back to
mechanism 1 rather than telling the user the topic is uncovered.
Then read the package. Open the exact
${CLAUDE_PLUGIN_ROOT}/skills/<domain>/<slug>/SKILL.md the lookup
returned, plus its references/gotchas.md and
references/llm-anti-patterns.md. Do not answer from this router:
it is a map, not the territory.
Domain routers
Hand off to the router for the domain the request lands in. Each one
carries that domain's featured skills, decision trees and templates.
salesforce-admin (261 skills) — Declarative Salesforce configuration: objects, fields, record types, page layouts, permission sets, reports, the record-access model (OWD, role hierarchy, sharing rules), and the requirements work that precedes them.
salesforce-agentforce (53 skills) — Agentforce and Einstein: agents, topics, actions, prompt templates, grounding, guardrails, evaluation and production readiness.
salesforce-apex (159 skills) — Apex and SOQL: triggers, Apex governor limits, async processing, OUTBOUND HTTP callouts, security enforcement, and test patterns. Owns calling an external API FROM Salesforce; salesforce-integration owns inbound. Generic nightly scheduling without naming code belongs to salesforce-flow. Codebase security review belongs to salesforce-security. NOT for SOSL — use salesforce-data.
salesforce-architect (106 skills) — Solution and platform architecture: multi-org strategy, scalability limits, licensing, Well-Architected reviews and architecture decision records.
salesforce-data (101 skills) — Data model, data movement and data quality: migrations, bulk loads, query optimisation, deduplicating at volume, archival. Ordinary-volume duplicate cleanup and prevention use salesforce-admin; come here for hundreds-of-thousands+ dedup or third-party tools. LDV architecture uses salesforce-architect.
salesforce-devops (70 skills) — Salesforce delivery: source tracking, packaging, branching, CI/CD pipelines, environment strategy and deployment troubleshooting.
salesforce-flow (63 skills) — Flow Builder: record-triggered, screen, scheduled and orchestration flows, bulkification, fault handling, limits, testing. "My flow" belongs here even when salesforce-apex also names the limit. Nightly scheduling without naming code defaults here; apex takes it when code/class/Apex is named. Flow-vs-Apex choice before anything is built: admin/process-automation-selection.
salesforce-integration (61 skills) — INBOUND integration and the API surface itself: the Salesforce REST and SOAP APIs, Bulk API 2.0 jobs, Platform Events, CDC, Pub/Sub, Named Credentials and middleware. For calling OUT to someone else's API from Apex, use salesforce-apex instead.
salesforce-lwc (83 skills) — Lightning Web Components: reactivity, wire adapters, component communication, accessibility, performance, security and Jest testing.
salesforce-omnistudio (34 skills) — OmniStudio: OmniScripts, FlexCards, DataRaptors, Integration Procedures, Business Rules Engine and DataPack deployment.
salesforce-security (49 skills) — Platform security and compliance: org hardening, encryption, session policy, MFA, monitoring, incident response, and TROUBLESHOOTING a specific record-access denial. Designing the sharing model itself (OWD, role hierarchy, sharing rules) is salesforce-admin.
Decision trees — read before choosing a technology
When a request straddles more than one technology, read the tree
before activating any skill, and cite the branch that decided it.
${CLAUDE_PLUGIN_ROOT}/standards/decision-trees/agentforce-capability-selector.md
${CLAUDE_PLUGIN_ROOT}/standards/decision-trees/async-selection.md
${CLAUDE_PLUGIN_ROOT}/standards/decision-trees/automation-selection.md
${CLAUDE_PLUGIN_ROOT}/standards/decision-trees/flow-pattern-selector.md
${CLAUDE_PLUGIN_ROOT}/standards/decision-trees/integration-pattern-selection.md
${CLAUDE_PLUGIN_ROOT}/standards/decision-trees/performance-tuning.md
${CLAUDE_PLUGIN_ROOT}/standards/decision-trees/sharing-selection.md
${CLAUDE_PLUGIN_ROOT}/standards/decision-trees/README.md — the index,
with the rule for when a tree outranks a retrieval tie
Cross-domain entry points
If you need a place to start before running a lookup, these are the
highest-traffic packages in the library:
${CLAUDE_PLUGIN_ROOT}/skills/admin/object-creation-and-design/SKILL.md — designing a new object or reworking an existing one
${CLAUDE_PLUGIN_ROOT}/skills/admin/permission-set-architecture/SKILL.md — who can see and do what
${CLAUDE_PLUGIN_ROOT}/skills/apex/trigger-framework/SKILL.md — any trigger work at all
${CLAUDE_PLUGIN_ROOT}/skills/apex/governor-limits/SKILL.md — anything that says "too many" or "exceeded"
${CLAUDE_PLUGIN_ROOT}/skills/apex/soql-fundamentals/SKILL.md — query correctness and selectivity
${CLAUDE_PLUGIN_ROOT}/skills/flow/record-triggered-flow-patterns/SKILL.md — the most common declarative automation
${CLAUDE_PLUGIN_ROOT}/skills/lwc/component-communication/SKILL.md — the first thing an LWC build gets wrong
${CLAUDE_PLUGIN_ROOT}/skills/data/data-model-design-patterns/SKILL.md — relationships and normalisation
${CLAUDE_PLUGIN_ROOT}/skills/integration/rest-api-patterns/SKILL.md — talking to or from an external system
${CLAUDE_PLUGIN_ROOT}/skills/security/org-hardening-and-baseline-config/SKILL.md — the org security baseline
${CLAUDE_PLUGIN_ROOT}/skills/architect/well-architected-review/SKILL.md — assessing an org against Salesforce Well-Architected
${CLAUDE_PLUGIN_ROOT}/skills/devops/salesforce-dx-project-structure/SKILL.md — getting metadata into source control
${CLAUDE_PLUGIN_ROOT}/skills/agentforce/agentforce-agent-creation/SKILL.md — building an Agentforce agent
${CLAUDE_PLUGIN_ROOT}/skills/omnistudio/omniscript-design-patterns/SKILL.md — OmniStudio guided flows
Run-time agents
For a whole workflow rather than a single question, invoke the
matching subagent. Each reads its full AGENT.md playbook, cites
every skill it consulted, and never deploys to an org.
- admin —
assignment-and-auto-response-rules-designer, audit-router, automation-migration-router, build-doc-keeper, build-planner, business-hours-and-holidays-configurator, change-impact-planner, config-workbook-author, csv-to-object-mapper, custom-metadata-and-settings-designer, email-template-modernizer, entitlement-and-milestone-designer, experience-cloud-admin-designer, field-impact-analyzer, fit-gap-analyzer, knowledge-article-taxonomy-agent, lead-routing-rules-designer, metadata-builder, object-designer, omni-channel-routing-designer, path-designer, permission-set-architect, plan-verifier, process-flow-mapper, profile-to-permset-migrator, requirements-clarifier, sales-stage-designer, salesforce-learning-guide, story-drafter, user-access-diff
- agentforce —
agentforce-action-reviewer, agentforce-builder, agentforce-quality-engineer
- apex —
apex-builder, apex-refactorer, apex-test-failure-triager, automation-transaction-profiler, security-scanner, soql-optimizer, test-class-generator, trigger-consolidator
- architect —
org-health-assessor-v2, salesforce-decision-facilitator, waf-assessor
- data —
data-loader-pre-flight, data-migration-reconciler, data-model-reviewer, duplicate-rule-designer
- devops —
build-step-runner, changeset-builder, deployment-failure-triager, deployment-risk-scorer, milestone-verifier, multi-org-drift-analyzer, release-readiness-reviewer, release-train-planner, sandbox-strategy-designer, step-tester
- flow —
flow-analyzer, flow-builder, flow-orchestrator-designer
- integration —
bulk-migration-planner, integration-catalog-builder, integration-incident-triager
- lwc —
lwc-auditor, lwc-builder, lwc-debugger
- omnistudio —
omnistudio-designer
- security —
access-path-explainer, security-posture-reviewer
Rules
- Never answer a Salesforce behaviour, limit or API question from
this file. Route, open the package, then answer from it.
- Cite every skill id, template path and decision-tree branch you
used. The library's value is that its claims are traceable.
- Never claim the library lacks coverage without having run a
lookup and pasted its output.
- Never deploy to an org and never write outside the paths the user
gave you.