solacedev/solace-developer-workshops/sam-author-agent
Claude Code and agent skill by solacedev/solace-developer-workshops
- community
Checking that the SKILL.md file is still in the repository.
Use when creating or editing Solace Agent Mesh agents or workflows — writing instructions/system prompts, enabling built-in tool groups, giving an agent an MCP server's tools, peer delegation between agents (inter_agent_communication), structured input/output schemas, agent cards/abilities, or multi-step workflow DAGs (switch/map/loop nodes). Not for creating connectors (sam-connectors), custom tools (sam-tools-and-skills), inbound entrypoints (sam-entrypoints), or choosing where SAM runs (sam-install-run / sam-deploy).
Install
npx vskill@latest install solacedev/solace-developer-workshops/sam-author-agentSKILL.md
sam-author-agent
This skill creates and edits SAM agents and workflows (behavior: instructions, tools, delegation, structured I/O, DAGs). For reaching external data/services no-code see sam-connectors; for custom tool code see sam-tools-and-skills; for making agents reachable from outside (Slack, MCP clients, events) see sam-entrypoints.
Agent or workflow? A single assistant that converses and uses tools = agent. A multi-step pipeline with branching, iteration, or fixed sequencing ("listen → filter → store", "research → summarize → send") = workflow (a DAG that orchestrates agents and tools). Workflows compose: entrypoints trigger them, connectors act for them.
Authoring paths — teach in this order
- Builder UI (default, especially for non-technical users). Creation lives under the WebUI's Builder nav group — not the Agents screen, which is read-only (running-agent cards, no create control). Two entry points: Builder → Quick Build (the AI-assisted chat builder: describe the agent/workflow in plain language, review the plan card, Build & Activate) and Builder → Agent Management (the manual path: Add Agent → Create New Agent → Create Manually, fill the sectioned editor, then Create and Deploy). Use the UI whenever every needed control exists in it (see capability map below).
- Declarative config —
sam config plan/applyvia the `sam-declarative-config` skill. The step up when a knob has no UI control, or the user wants version-controlled, repeatable config. That skill is the only source of full YAML blocks — it regenerates from the live schemas. This skill names knobs and keys (safe to mention to the user); writing out a complete config block, nesting, or any key this skill doesn't name issam-declarative-config's job. Never bridge the gap from memory. - Raw runtime YAML — escape hatch only (debugging, parity checks). Never the headline.
Builder UI capability map (what exists — verified 2026-06)
Name & description · instructions/system prompt · model alias selection only (no temperature/tuning) · toolsets with per-tool exclusion and per-tool config · skills (built-in + custom) · connectors (incl. MCP, via the connector wizard) · agent card abilities (≤10), input/output modes · deploy on save. Workflow builder: visual DAG with agent / workflow / switch / map / loop nodes, per-node settings, workflow-level timeouts/retry/exit handlers.
No UI control exists for (go straight to declarative config; say so plainly — never invent a UI control):
- Peer delegation (
interAgentCommunication) — see references/peer-delegation.md - Structured input/output schemas — see references/structured-output.md
- Welcome message + suggestion chips, required scopes (RBAC), discovery/publishing tuning, and the advanced behavior knobs (budgets, streaming, artifact handling…) — all API-reachable via the agent's
additionalConfigurations; author through declarative config. - Model tuning parameters (temperature, top_p, …) are not even API-reachable — runtime YAML only. Flag this honestly when asked.
Hard rules (each counters an observed failure)
- Go product, Go patterns. Never use Python SAM shapes (
apps:/app_config:files, thesolace-agent-mesh-plugin-creatorskill, ADK idioms). Runtime YAML is compatible with Python SAM, but it is the escape hatch, not the taught path — and "compatible" is not a license to write its keys from memory. - Check the built-in catalog before recommending external tools. Web search is built in (
web_search_googlein theweb_toolsgroup) — do not send users to install a search MCP server for plain web lookup. See references/builtin-tools.md. - MCP server on an agent has a UI path: the MCP connector wizard (URL, auth, tool selection). Lead with it. See references/mcp-on-agent.md.
- When unsure of a key name, you are not allowed to guess — invoke
sam-declarative-configand look it up.
References (Drew-depth detail)
| Topic | File | |
|---|---|---|
| Built-in tool groups — canonical names, what's in each, UI toolsets vs runtime groups | references/builtin-tools.md | |
| Giving an agent an MCP server's tools (UI / config / runtime paths, auth, filtering) | references/mcp-on-agent.md | |
| Peer delegation (multi-agent hand-off) | references/peer-delegation.md | |
| Structured input/output (agent as typed function) | references/structured-output.md | |
| Workflows — node types, settings, composition | references/workflows.md |
Skill bundles (packaging behavior for reuse) are attached here via the builder's skills picker, but creating them belongs to sam-tools-and-skills.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingsample_configuration/.claude/skills/sam-author-agent/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/solacedev/solace-developer-workshops/sam-author-agent)