steloit/squad-skills
4 skills4 installsGitHub
Install
npx skills add https://github.com/steloit/squad-skillsSkills in this repo
1SquadSquad is a journey-wide agent workflow from Steloit that coordinates multiple specialized agents—refiner, planner, critic, builder, shield, inspector, and ranger—through a safety-first, card-driven pipeline for solo and indie builders shipping with Claude or Codex-class models. Default provider configuration maps each role to Opus, Sonnet, or GPT-5.x variants and sets higher reasoning effort on planner and builder steps so planning and implementation stay deliberate rather than fast-but-wrong. Before any card moves forward, Squad demands codebase-first exploration: read relevant files, map interfaces, and follow existing patterns instead of guessing architecture. Cards that are too large, ambiguous, or unscoped against real code should split or return to squad-refine. Use it whenever you want agent squads to mirror a small team's refine-plan-review-build-ship discipline across validate scoping, build implementation, and ship security review—not a single chat thread improvising structure.1installs2Squad Exploresquad-explore is a squad-pipeline agent skill for when you know the problem but not the implementation path. It receives a topic (or runs a short clarification interview if context is thin), explores the codebase deeply, and produces a direction report plus phased tasks on the squad board—explicitly stopping before code changes. Solo builders using multi-agent squads benefit because it separates reconnaissance from execution, aligns with mandatory safety principles, and avoids half-implemented guesses. Pair it when uncertain direction would otherwise waste tokens on wrong files or conflicting refactors. It integrates with the broader squad shared config, pipeline levels, and API endpoints documented in sibling squad files. Not a drop-in for “just fix this bug” requests; use direct implementation skills when the path is already obvious.1installs3Squad Refinesquad-refine is an agent skill for solo builders using the Squad task pipeline who need disciplined backlog grooming without skipping safety context. Invoked as `/squad-refine <ID>`, it loads the card from the project API, insists on todo status (with a user warning otherwise), then runs a structured interview that converts a rough title and description into actionable requirements: explicit goal, bounded scope, acceptance criteria, and edge cases. Before interviewing, it hunts dependency hints in description and tags, pulls prior card implementation notes and plans, and inspects the codebase when a predecessor exists; if none is named, it asks one upfront question about prior work. The procedure is wired to shared pipeline levels, status transitions, and API endpoints documented in squad/shared.md, and it treats squad/principles.md as mandatory safety reading—not optional flavor text. Use it when backlog items are too fuzzy for an implementation agent and you want repeatable refinement inside the same repo workflow.1installs4Squad Runsquad-run is the orchestration skill for Steloit’s squad system: it runs tasks through a seven-column board using six specialized agents while keeping status transitions and API-backed state consistent. Solo builders install it when they want a repeatable factory line—plan, critique, build, harden, inspect, and range—rather than letting the main agent improvise implementations that drift from board rules. Auto-trigger rules are strict: any implement task NNN pattern, continuations like next task, or a short confirmation after Claude names a task ID should invoke /squad-run instead of silent coding. The skill pairs with shared squad.md for pipeline levels, endpoints, and error handling, so it behaves like production PM for agent teams. It is most valuable when you already track work as numbered squad tasks and need review and safety roles baked into every pass.1installs