solacedev/solace-developer-workshops/sam-author-agent

Claude Code and agent skill by solacedev/solace-developer-workshops

Trust tier 3
  • community
Source check pending

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-agent

SKILL.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

  1. 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).
  2. Declarative config — sam config plan / apply via 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 is sam-declarative-config's job. Never bridge the gap from memory.
  3. 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, the solace-agent-mesh-plugin-creator skill, 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_google in the web_tools group) — 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-config and look it up.

References (Drew-depth detail)

TopicFile
Built-in tool groups — canonical names, what's in each, UI toolsets vs runtime groupsreferences/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, compositionreferences/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

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
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/solacedev/solace-developer-workshops/sam-author-agent/badge)](https://verified-skill.com/skills/solacedev/solace-developer-workshops/sam-author-agent)

Something wrong with this skill or its page? Tell us.