メニュー
Write step-by-step automation instructions for a workflow, tailored to your tool (Claude, n8n or Zapier). Best run right after discover-patterns or process-analyst.
(new) Find a person's repeated tasks from their screen activity, and what each one is worth automating, with the time and money saved. Outputs the numbers and data a report is built from, not the visual. Use to analyze processes, mine workflows, or see what is automatable and what it saves. Every figure is labelled and grounded, never made up.
(new) Turn a process analysis into a polished, client-ready report, first an HTML deck you review, then a matching PDF. Use to package an analysis into an exec report or PDF. Shows the deck for approval before making the PDF, and never makes up numbers.
Generate a time report grouped by client/project
Discover repeated workflow patterns from screen activity and suggest automations
Summarize what you've been doing in the last 30 minutes.
Prepare a MemoryLane release by updating the version and release notes, then creating and pushing the tagged release commit that triggers CI. Use when the user asks to release, ship, publish, bump version, or cut a stable or prerelease version.
| name | automation-instructions |
| allowed-tools | AskUserQuestion, mcp__memorylane__browse_timeline, mcp__memorylane__search_context, mcp__memorylane__get_activity_details |
| description | Write step-by-step automation instructions for a workflow, tailored to your tool (Claude, n8n or Zapier). Best run right after discover-patterns or process-analyst. |
Turn one workflow into clear, build-ready instructions for a specific automation tool.
Assumes it runs after /discover-patterns or /process-analyst-new, so it reuses
their output as the source and does NOT re-mine the timeline unless nothing is supplied.
In priority order, do NOT re-derive what you already have:
/discover-patterns (a selected pattern + its details) -> use it as-is./process-analyst-new (a process-analysis*.json ledger, or a process
the user points to) -> read that process's steps, apps, frequency, and automatable split.
Glob process-analysis*.json in the working dir; if found, use it.Only if NONE of the above is available, fall back to mining: search_context(query) across
~30 days for instances, browse_timeline around the strongest matches for the full sequence,
and get_activity_details(ids) on 2 to 3 clear instances for exact steps. Keep it bounded.
Respect the constraints everywhere in Step 3, and call out any step the constraints force to stay human.
Reconstruct the flow from trigger to done, separating variables (change each run, the inputs) from constants (fixed, hardcoded). Then write it for the chosen tool:
Render in the chosen format: markdown = the skeleton below; text = the same content
as plain prose, no markdown; JSON = one object { name, tool, trigger, variables[], constants[], steps[{n, app, action, input, output, on_error}], human_steps[], constraints[] }.
Markdown skeleton (keep it tight, drop empty sections):
# [Workflow] - Automation instructions ([tool])
**Does:** [one line] **Trigger:** [what starts it] **Frequency:** [from the source]
**Constraints:** [what it must not do]
## Variables (inputs that change each run)
- [name]: [example]
## Constants (hardcoded)
- [name]: [value]
## Steps
1. [Action] in [app] - input [x] -> output [y]. On error: [fallback].
2. ...
## Stays human
- [step + why, incl. anything the constraints forced manual]
## Build notes ([tool])
- [trigger + key apps/nodes/actions + effort: easy/medium/hard]
Ask where to save (default ~/Desktop/automation/[workflow-slug].md, or .txt / .json
to match the format). Create the folder if needed. Then show the path and a 2-line summary
(tool, step count, key apps). Offer to do another workflow.