Agentforce Agent Script
End-to-end Salesforce Agentforce agent lifecycle: design, Agent Script, validate, deploy, publish.
Test report
- Verdict
- Tested · Works
- Score
- Tested
- Jul 13, 2026
- Environment
- Claude Code 2.x (agent harness)
- Upstream re-checked
- Aug 10, 2026 · 2ce233b
Installed with a command from a clean HOME without ~/.claude — SKILL.md was in place; all ~30 references to references/ and assets/ in the body resolved, no scripts or network calls in the skill (Stage 0 clean). Since the last test, the author added an entirely new domain 'Optimize an Agent' (v0.8 → v0.9) with four patterns and included 'optimize/improve/refactor' in the description — the old finding about Rule 8 (else-if prohibition) is still there and untouched. Ran the new domain on my .agent-file with four injected defects: without the skill, I invented non-existent syntax `@actions.fetch_case.outputs.caseRecord` (core-language explicitly states: `@outputs` lives only in `set`/`if` immediately after an action), left `@variables.customerName` without `{!...}` and all conditional logic in prose; with the skill — correct relay via a mutable variable, 4 `set @variables.X = @outputs.Y` bindings, `{!...}`-injections, and flat `if` statements according to Rule 8. Minus: new pattern files and the report template in SKILL.md require `orderRecord: mutable object` and `roomAvailable: boolean` without defaults, which contradicts the skill's own core-language reference (section 6 and Anti-Pattern 12: mutable variables MUST have a default value) — for this, points were deducted from output and docs. Could not check the 'sf agent validate' step: CLI `sf` and a Salesforce org were not in the environment, compilation was not measured — only checked against the skill's own documented rules.
Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works
- Installs cleanly 5/5
- Triggers reliably 5/5
- Output vs. baseline 9/10
- Docs & honesty 4/5
What Agentforce Agent Script does
Routes the full Agentforce agent development lifecycle — Agent Spec design, writing .agent files in Agent Script, sf CLI validate/preview/deploy/publish/activate, and behavioral debugging via session traces — through 18 detailed reference files. Triggers on .agent file work, aiAuthoringBundle metadata, subagent/action design, and sf agent CLI commands; explicitly excludes Apex, Flow, and Experience Cloud work.
How to install Agentforce Agent Script
git clone https://github.com/SalesforceAIResearch/agentforce-adlc
cd agentforce-adlc
mkdir -p ~/.claude/skills
cp -r skills/agentforce-generate ~/.claude/skills/agentforce-generate
Skills live in ~/.claude/skills/ (global) or .claude/skills/
(per-project). Restart Claude Code after installing.
Commands — how to trigger Agentforce Agent Script
-
/agentforce-generateEnd-to-end Salesforce Agentforce agent lifecycle: design, Agent Script, validate, deploy, publish.
It also activates on plain-language prompts like these:
-
Add a billing subagent to our Agentforce agent for unpaid invoices -
Write an Agent Spec and design the subagent graph for a support agent -
My sf agent validate is failing, help debug this .agent file
Frequently asked questions
- Is the Agentforce Agent Script skill free?
- Yes. The skill itself is free from SalesforceAIResearch/agentforce-adlc. SkillProof publishes the install command and an independent test verdict at no cost.
- Does Agentforce Agent Script work with Claude Code?
- We tested it with Claude Code 2.x (agent harness) on Jul 13, 2026. Verdict: Tested · Works. Installed with a command from a clean HOME without ~/.claude — SKILL.md was in place; all ~30 references to references/ and assets/ in the body resolved, no scripts or network calls in the skill (Stage 0 clean). Since the last test, the author added an entirely new domain 'Optimize an Agent' (v0.8 → v0.9) with four patterns and included 'optimize/improve/refactor' in the description — the old finding about Rule 8 (else-if prohibition) is still there and untouched. Ran the new domain on my .agent-file with four injected defects: without the skill, I invented non-existent syntax `@actions.fetch_case.outputs.caseRecord` (core-language explicitly states: `@outputs` lives only in `set`/`if` immediately after an action), left `@variables.customerName` without `{!...}` and all conditional logic in prose; with the skill — correct relay via a mutable variable, 4 `set @variables.X = @outputs.Y` bindings, `{!...}`-injections, and flat `if` statements according to Rule 8. Minus: new pattern files and the report template in SKILL.md require `orderRecord: mutable object` and `roomAvailable: boolean` without defaults, which contradicts the skill's own core-language reference (section 6 and Anti-Pattern 12: mutable variables MUST have a default value) — for this, points were deducted from output and docs. Could not check the 'sf agent validate' step: CLI `sf` and a Salesforce org were not in the environment, compilation was not measured — only checked against the skill's own documented rules.
- What is the Agentforce Agent Script SkillProof Score?
- 9.2/10 — installs cleanly 5/5, triggers reliably 5/5, output vs. baseline 9/10, docs & honesty 4/5.
- How do I install Agentforce Agent Script?
- Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.
- Can I use Agentforce Agent Script with Cursor, Copilot, Gemini CLI, Codex or other AI tools?
- The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.