Poteto Mode
An agent skill by cursor, from cursor/plugins. Tags: development, quality, review, workflow, writing.
What it does
poteto's agent style for concise, detailed responses, deliberate subagents, unslopped prose, simple code, and verified work. Use for poteto, /poteto-mode, or requests to work in this style.
Install
With the skills CLI, which installs into Claude Code, Codex, Cursor and other agents:
npx skills add cursor/plugins --skill 'Poteto Mode'Or copy the skill folder into Claude Code's skills directory by hand (~/.claude/skills for every project, or .claude/skills inside one):
git clone --depth 1 https://github.com/cursor/plugins
cp -r plugins/pstack/skills/poteto-mode ~/.claude/skills/'Poteto Mode'Safety box score
Grade D, elevated risk: 58 of 100 risk points, rated 2026-10-16. The grade rates what the skill and its scripts can reach on the machine of whoever installs it, across eight categories from shell execution to secrets access. How the score works.
Poteto Mode is a workflow skill for agent autonomy, PR watching, shipping, worktree cleanup, and plan validation. Its main risks are broad write powers over GitHub branches and local worktrees, plus dependency installation and access to local agent transcripts outside the repo. Some supporting implementation files were not readable in this review, so a few behaviors remain unconfirmed.
| Category | Level | Why |
|---|---|---|
| Code & shell execution | 2 (moderate) | `scripts/bootstrap.ts` calls `Bun.spawnSync([process.execPath, "install", "--frozen-lockfile"])` and then re-execs itself, so the skill can run shell commands and restart the process. |
| Network access | 2 (moderate) | The bootstrap path can fetch dependencies during `bun install`, and the playbooks drive `gh` or `origin pr` operations against remote GitHub state. |
| Filesystem reach | 3 (high) | `playbooks/worktree-cleanup.md` allows `git worktree remove --force <path>` and `rm -rf`, and `scripts/worktree-audit.sh` reads `~/.cursor/projects/.../agent-transcripts` outside the project. |
| Secrets & credentials | 1 (low) | The skill does not target credential stores, but it does read local agent transcripts under the user's home directory, which may contain sensitive content. |
| Privilege & persistence | 0 (none) | I found no `sudo`, global install, service registration, shell startup edit, or OS-level privilege requirement in the readable files. |
| Instruction hijack surface | 2 (moderate) | `SKILL.md` and the playbooks explicitly tell the agent to bypass human confirmation for reversible work and to keep going under stop phrases like `run until done`, which suppresses normal pause points. |
| Irreversible actions | 3 (high) | `playbooks/shipping.md`, `playbooks/autopilot-full.md`, and related files authorize merging, retargeting, force-pushing, deleting, and pruning branches or worktrees, which is direct remote and local action. |
| Transparency | 2 (moderate) | Key support files were not readable in this review, including `scripts/watch-pr/github.ts`, `scripts/watch-pr/policy.ts`, and the `scripts/orch/*` files, so some behavior could not be audited. |
Before installing
- Read `scripts/bootstrap.ts` first. It installs dependencies and restarts the process.
- Read `playbooks/worktree-cleanup.md` before using cleanup flows. It can remove worktrees and caches.
- Inspect the unread `scripts/watch-pr/github.ts` and `scripts/watch-pr/policy.ts` files before trusting the PR watcher.
- Run the skill in a scratch clone first if you do not want it touching `~/.cursor/projects/.../agent-transcripts` or local worktrees.
Rated 2026-10-16 by an automated review of 38 files from the skill's repository.
Source
- Repository
- cursor/plugins (all skills from this repository)
- Path
- pstack/skills/poteto-mode/SKILL.md
- Branch
- main
- Collection
- cursor
- Updated
- 2026-09-23
Related skills
- unslop — Cut AI tells from any writing. Must always apply.
- deslop — Remove AI-generated code slop and clean up code style
- maintain-verification-skill — Periodic pass that keeps a project's verification skill and feature map honest: parallel source readers per feature, one live session driving every feature.
- ralph-loop — Start a Ralph Loop for iterative self-referential development.
- review-and-ship — Review the current branch for bugs, intent fit, and test coverage; run or write tests; commit focused work; open or update a PR.
- show-me-your-work — Keep a reviewable decision trail for long-running or unattended work: a TSV log with one row per decision (what, why, evidence, result).