Wraps an API into a Cursor/Agent Plugin (MCP + skills). Data shape first, smallest scaffold that works, prove locally, then ask once for affiliation and publish to Marketplace or cursor.directory.
When wrapping an API as a Cursor/Agent plugin: prefer Agent Plugin (root plugin.json + skills + optional mcp.json) unless rules/hooks/agents are required; name the data shape first; keep public no-auth HTTP MCP servers zero-dep stdio until an SDK earns its install; never publish to marketplace unless the owner explicitly says to.
The assistant's capabilities include a plugin named 'pstack'.
The assistant operates with a 'Poteto Mode' style when invoked, emphasizing concise, detailed, and verified work. This includes starting multi-step tasks with principle-driven todolists, and applying principles like 'prove it works', 'model the domain', 'laziness protocol', and 'foundational thinking' to guide decisions and actions, citing which principle shaped specific choices.
Publish routing: ask once if the owner works at the company/service — yes or they insist → Cursor Marketplace; no/unsure → cursor.directory. Trust their answer; the approve gate still blocks submit.
Grok Bot does not support loading local plugins from ~/.cursor/plugins/local. It loads plugins only from the Cursor dashboard/marketplace. Proving local install still matters for Cursor IDE; note the gap when verifying on Grok Bot itself.
2
Skills
From the Marketplace
wrap-api-as-plugin
Use when wrapping an HTTP API, OpenAPI/spec, or "wrap X" into a Cursor/Agent Plugin (MCP + skills). Prefer before scaffolding files.
prove-plugin-local
Use when proving a Cursor/Agent Plugin loads locally — schema validate, install under ~/.cursor/plugins/local, and record what loaded.
publish-cursor-plugin
Use when publishing a Cursor/Agent Plugin: ask once if they work at the company/service — yes → Marketplace, no/unsure → cursor.directory. Trust their answer; approve gate still blocks submit.
if you want to go fast, go deep first. pstack helps you write less, but higher quality code. rigorous agent workflows you can parallelize with confidence.