← ClayCONTENT HISTORY

Update to Clay

Snapshot Nov 18, 2026 · 06:19 UTC · version 2.38.0

Collection source: downloaded plugin package.

WHAT CHANGED

First saved snapshot

No earlier snapshot is available to establish a change.

Save Clay and check future changes in Following.

Up to 3 plugins free · no card required.

Compare saved observations

Download comparison JSON
Full technical diff · 0 changed fields
Full snapshot data
{
  "description": "Clay workflows — simplify a workflow via the CLI (`clay workflows` commands): merge redundant nodes, cut unnecessary complexity, and replace LLM nodes with deterministic alternatives where possible.",
  "included_files": [],
  "name": "workflows-simplify",
  "skill_md_contents": "---\nname: workflows-simplify\ndescription: 'Clay workflows — simplify a workflow via the CLI (`clay workflows` commands): merge redundant nodes, cut unnecessary complexity, and replace LLM nodes with deterministic alternatives where possible.'\n---\n\n# Simplifying a workflow\n\nBefore using this skill, read the `clay` entry skill if you haven’t already this session, and follow its access guidance.\n\nAnalyze the current workflow and suggest concrete simplifications to reduce complexity, improve reliability, and lower costs.\n\n## Process\n\n1. **Read the workflow** using `clay workflows graph get <workflowId> --mode full` to get all node details\n2. **Analyze each node** against the simplification checklist below\n3. **Present findings** as a prioritized list of suggestions with specific changes, alongside a render of the **current graph** (`clay workflows diagram <workflowId>`) so the user can see which nodes each suggestion affects\n4. **Apply authorized changes** — edit the workflow only when the user's request authorizes modifications, via `clay workflows nodes update`/`create`/`delete`. Once authorized, apply clearly behavior-preserving improvements as you go and ask before changes with a material behavior or quality trade-off\n5. **Show the result** — after applying, run `clay workflows graph format <workflowId>` and render the **updated graph** so the simplification is visible, not just described\n\nNarrate throughout and prefer the diagram over raw node JSON — see the `workflows` skill's `presenting.md`.\n\n## Simplification Checklist\n\n### Replace LLM nodes with code nodes\n\nRegular (LLM) nodes cost an LLM call per execution. Many can be replaced with deterministic code:\n\n- **Data transformation** — extracting fields, reformatting JSON, string manipulation → code node\n- **Simple routing** — if the decision can be expressed as rules on data fields → conditional node (rules mode)\n- **Calculations** — math, aggregation, counting → code node\n- **Template filling** — constructing strings from known fields → code node\n\n**Ask:** \"Does this node require creative reasoning, or could a Python function do the same thing?\"\n\n### Merge sequential nodes\n\nTwo adjacent nodes can often be combined into one if:\n\n- Node A passes all its output to Node B, and Node B doesn't add new tools or branching\n- Both nodes use the same model and could be described in a single prompt\n- One node just reformats the other's output\n\n**Ask:** \"Would combining these prompts into one still produce the same result?\"\n\n### Remove unnecessary nodes\n\n- Nodes that just pass data through without transformation\n- Conditional nodes with only one possible outcome\n\n### Pin typed inputs deterministically\n\nWhen a downstream node needs specific typed data from an upstream node:\n\n- Add `outputSchema` to the upstream node\n- On the downstream agent or code node, pin each input via `sourceNodeId`/`sourcePath` inline on the `inputSchema` property (see the `workflows` skill's `data-passing.md`)\n- This preserves exact values across nodes\n\n### Replace a threaded state object\n\nLook for code nodes that each take one serialized state input (often `state_json`) from the previous code node and return it changed. That state gets copied into every step it passes through, so step inputs and outputs keep growing. Suggest:\n\n- A declared `outputSchema` on each code node with the fields it produces\n- Downstream nodes pinning only the fields they read, from the node that produced them\n- One final code node that builds any combined summary from the stage outputs\n\n### Simplify tool usage\n\n- If a node has tools it never uses, remove them (reduces prompt size and cost)\n- If a node calls one Clay action and passes the result, make it a tool node rather than an agent node\n\n### Use code mode for conditional nodes\n\n- Conditional nodes: prefer `rules` or `code` mode over `agentic` mode when the decision logic is expressible programmatically\n\n## Output Format\n\nPresent suggestions as:\n\n1. **What to change** — specific node(s) affected\n2. **Why** — what complexity or cost this removes\n3. **How** — the concrete edit (new node type, merged prompt, code snippet)\n\nPair the suggestion list with the current-graph render so each affected node is easy to locate. For analysis or recommendation requests, present the suggestions without editing. If the user asks you to modify the workflow, apply clearly behavior-preserving improvements as you identify them and state your assumptions. Ask only when an edit has a meaningful behavior or quality trade-off. Then run `clay workflows graph format <workflowId>` and show the updated graph so the user can see the before/after difference.\n"
}

SHA-256 of public snapshot: fe4c1e585193b71d42e2abaed04d341ab94621f76a0918edd7727fce4eace75d