Agentforce Architecture Analyze

Pulls a complete design-time snapshot of an Agentforce agent: planner, topics, actions, flows, Apex classes, prompt templates, and NGA plugins. You get a human-readable architecture doc and a Mermaid invocation graph, all from declared metadata, not runtime audit rows. The pipeline runs inline with parallel Tooling SOQL queries to keep typical runs under 45 seconds, though large bots with many flows will scale linearly. Good for inventory audits, version diffs, or just understanding what an agent is actually wired to do before you touch it. Point it at an org alias and agent API name, and it writes everything under a versioned output directory you can diff or archive.

skills.sh installs: 5.9K
GitHub stars: 1.2K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add forcedotcom/sf-skills --skill agentforce-architecture-analyze --agent claude-code

Installs into .claude/skills of the current project.

Files

agentforce-architecture-analyze — declared architecture snapshot

Design-time metadata tree for one Agentforce agent: planner → topics → actions → flows → Apex → prompts → NGA plugins. Reads declared metadata only — BotDefinition, GenAiPlanner*, GenAiPlugin*, GenAiFunction*, Flow, ApexClass, GenAiPromptTemplate. Does not read runtime audit rows.

Runtime budget: 30–45s typical, ≤60s hard cap on reference fixtures. Sequential baseline would be 90–220s; parallel Tooling SOQL fan-out delivers a 3–5× speedup. Large bots with many flows scale approximately linearly — each flow metadata retrieve is one round-trip.

Runs inline — no subagent. Every phase is deterministic file processing.

If the user hasn't named an agent

Trigger condition: $ARGUMENTS names no agent (no --agent flag and no exact BotDefinition API name in the prose). Never invent an agent name. Instead, discover what the org actually contains — one read-only query, no pipeline yet. Use the --org alias if given, otherwise the default target org (omit --target-org):

SKILL.md on GitHub