Copy every plugin from a Cursor marketplace catalog into /.cursor/plugins/local as real files (no symlinks). Use when installing or refreshing wild-horses (or any dual-manifest marketplace) for Cursor, when team marketplace refresh is unavailable, or when local plugin installs must not point at a checkout via symlink.
Apply the plain-language writing standard to prose anywhere in a repo: code comments, docstrings, markdown, plain text, or pasted text. /plain-language:check reports long sentences, em-dashes, filler phrases, and figurative tokens; /plain-language:apply rewrites until the scan is clean while a bundled stdlib CLI…
Rewrite prose to the plain-language standard until the scan is clean, and prove no code changed. Covers code comments, docstrings, markdown, plain text, a PR's changed lines, the branch diff, or pasted text.
Scan prose for plain-language violations and report them without editing. Covers code comments, docstrings, markdown, plain text, a PR's changed lines, the branch diff, or pasted text.
A local task-management system built around plans in a /plans/ / tree: a task is a plan that either spawns more plans or gets built into code. Capture from conversation, route to the next step, archive, dispatch via groundcrew, and optionally export to Linear/Jira.
Manage the groundcrew dispatch queue - view it, queue plans for pickup, promote plans to todo in bulk, or dequeue them. Shows the current repo's queue by default; show every repo with "all repos". Use when the user asks to see or manage the groundcrew queue, queue a plan for groundcrew, promote plans to todo, dequeue…
List saved plans and route the picked one to the right next skill, based on whether it reads as an idea, spec, or execution-ready plan. Use when the user asks to work on a saved plan, do a plan, implement a plan, execute a plan, pick up a plan to work on, or resume a plan from disk.
Archive a completed plan into /plans/ /done/ with a completion stamp. Use when the user finishes a plan, marks a plan done, archives a plan, says "I'm done with the plan", or asks to clear a completed plan.
File a plan as a Jira ticket, or update an existing one from a plan; supports first-time Jira setup inline. Use when the user asks to push a plan to Jira, file a Jira ticket/issue from a plan, or update an existing Jira ticket from a plan.
File a plan as a Linear ticket, or update an existing one from a plan; supports first-time Linear setup inline. Use when the user asks to push a plan to Linear, file a Linear ticket/issue from a plan, or update an existing Linear ticket from a plan.
List the saved plans for a repo (or a named bucket) as a read-only inventory of what's in /plans/ /, without picking one to work on. Use when the user asks to list, show, or see the saved plans for a repo.
Save a plan, design doc, or any file (markdown, JSON, YAML, etc.) into the per-repo /plans/ / tree, including paired outputs like task-list-builder's .json + .md pair. Use when the user asks to save a plan, save the plan, persist the plan, capture planning notes for future reference, or store a plan for later.
Split one plan into several smaller, independently-dispatchable plans wired with native Blocked-by dependencies, saved into the per-repo plans tree. Use when the user asks to split a plan into independently-grabbable pieces, decompose a plan/spec/design into child implementation tasks, or break a plan into vertical…
Edit a plan's frontmatter (status, assigned agent, and other fields) via plankeepercli.py file-meta set. Use when the user asks to change the agent on a plan, promote a plan to todo, mark a plan as ready, set a plan's status, or edit plan frontmatter.
Prints your branch's pull-request link the moment gh knows it. A run that spends twenty minutes in CI can no longer finish without printing that link. Also reports at every turn end whether the last commits were actually pushed, and whether the working tree is dirty. Everything comes from real git/gh state, never fro.
Run pyright on a Python codebase and fix the errors it finds via the harness task-list pipeline. Phase 3 partitions errors per-file into a structured markdown report; Phase 4 hands off to the sibling task-list-builder + task-list-runner skills for serial, audited fixes with per-task pyright verifySteps and inten.
Run pyright on the Python code in this project and systematically fix the errors using documented fix patterns. Supports optional strictness override (basic/standard/strict), persisting the level to config, progressive ratcheting, scoped runs, fix-intent selection, and an optional "suggested improvements" section at…
Argue the strongest good-faith case AGAINST the proposed changes in the current conversation or a named design/plan file. A built-in red-team voice: surfaces hidden costs, wrong assumptions, simpler alternatives, and the do-nothing option, so a plan gets stress-tested before it ships.
Argue the strongest good-faith case AGAINST the proposed changes. Builds a steelman of the opposition — hidden costs, wrong assumptions, simpler alternatives, second-order effects, and the do-nothing option — for a plan, design, or diff under discussion. Use when the user says "steelman the case against this", "argue…
Pull every configured git repo from origin in one shot, using git fetch + git merge --ff-only so diverged histories never auto-merge silently. Maintains a repo list at /.config/wild-horses/update-git-repos/repos.json, supports bootstrap auto-discovery under a root dir, manual add/remove, and a configurable default.
Pull every repo listed in /.config/wild-horses/update-git-repos/repos.json from origin/ via the bundled CLI, including bootstrapping that config (auto-discovery under a root directory, plus manual add/remove). Use when the user asks to update all their git repos, pull main on all repos, sync repos from origin, refresh…
Open a pull request with an architecture-first description (via its bundled summary-writer skill), review it, and babysit CI and review feedback. Ships /wild-pr:review for rubric-based code review, /wild-pr:babysit for single-pass tending, /wild-pr for create-then-babysit, and the summary-writer skill for architecture.
Watch a PR through CI and review feedback. Auto-fix high-confidence failures and address review comments. Use when the user says /wild-pr:babysit, babysit a PR, or respond to PR comments.
Code review of a diff, branch, or PR, with findings posted as anchored PR comments. Use when the user asks to review a diff, branch, or PR, asks to check a change against its ticket/spec/PRD, or runs /wild-pr:review [pr-number-or-url] [--effort low|high] [--report].