Drafts a personal -mode skill by mining agent transcripts and user preferences.
npx skills add https://github.com/edivad1999/stuc-stack --skill automate-me-edivad1999
Please help me install this Agent Skill. Skill: automate-me Source: https://github.com/edivad1999/stuc-stack/tree/main/skills/automate-me Command: npx skills add https://github.com/edivad1999/stuc-stack --skill automate-me-edivad1999
-mode skill (e.g. jay-mode) tailored to your working style. ## Core Features & Use Cases - Transcript Mining: Runs parallel subagents across slices of your workspace's recent agent history to surface recurring preferences with evidence pointers, cross-checking signals across slices before codifying them. - Structured Preference Capture: Uses multi-choice AskQuestion rounds plus one open question to capture intent that history mining misses. - Guided Authoring and Iteration: Orchestrates Cursor's create-skill for authoring and the unslop skill for prose discipline, placing the result in docs/<handle>-mode/SKILL.md with a thin harness delegate, then landing it via worktree and PR. - Use Case: You notice agents keep ignoring your conventions. Say "automate me" and the Skill mines your last few weeks of sessions, confirms findings with you, and produces a you-mode skill that future agents load on demand. ## Quick Start Tell the agent "automate me" or "create my -mode skill" to start the guided flow that mines your history and drafts your personal mode skill.High-intent search queries and answers about installing and using this skill.
Invoke the automate-me flow by saying "automate me" or "create my -mode skill". It mines your recent agent transcripts for recurring patterns, asks structured questions to fill gaps, then drafts a personal mode skill using create-skill and unslop.
Parallel subagents each read a slice of your workspace's agent-transcripts directory and return structured lists of patterns with evidence. Signals seen in two or more slices are treated as high-confidence; lone signals are usually dropped to avoid overfitting.
Project-local skills go to docs/<handle>-mode/SKILL.md with a thin harness delegate in .cursor/skills/, .claude/skills/, or .codex/skills/. User-global skills go to the harness home directory such as ~/.cursor/skills/<handle>-mode/.
Yes. The flow first searches for existing *-mode/SKILL.md files and defaults to update mode on repeat runs. It mines only history since the skill's last git edit and revises sections with new evidence while preserving uncontradicted content.
Skip it for task-specific skills or single narrow workflows like commit message conventions; use create-skill alone instead. Mode skills are for broad working conventions, not one-off procedures.
jnyrossDrafts a personal mode skill by mining agent transcripts and user preferences.
shalevaminStage, commit, and push code changes to a new branch and open a draft GitHub pull request.
stevejbickleyReviews academic prose for argument, structure, clarity, and consistency without rewriting.
andersnygaardGenerates a single actionable task file for your focused feature or idea.