agent-tools

Backlog-scale orchestration on top of /workflow — drives PM items through refine → plan → implement → review → local-merge via role-specialized sub-agents dispatched in parallel waves, each item isolated in its own git worktree. Bare /swarm summarizes state; /swarm <goal> runs the orchestrator; /swarm:setup sets up the charter and .agent-tools/ umbrella; /swarm:continue resumes a paused run.

Capture durable knowledge from any engineering work — debugging solutions, refactors, features, design decisions, reusable patterns — and maintain memory quality so each unit of work compounds the next

Principal workflow entry — resolve portfolio mode (swarm resume/handoff or unit state machine), drive work without inventing path; hard-stop when path is not established

Initialize and maintain planning root (.agent-tools/planning preferred), conventions (tracks, gates, integration), runs ledger scaffold, and shared memory under .agent-tools/memory/ (AGENTS.md memory-link) for the workflow family.

Parent skill for the workflow family — principal session entry is /workflow:continue (also bare /workflow). Horizon mapping, brainstorm, refine, plan, execute, review, audit, compound. Vertical-slice and deliverable-partition modes.

Detect gaps in the canonical skill corpus, propose targeted improvements, validate changes, and present for review. Inventories src/** skill trees; optional run-ledger seeds for process IP gaps.

Meta-skill responsible for the health, growth, and portability of the canonical skill corpus. Provides capabilities for importing skills from other agent platforms and evolving existing skills over time.

Flexible code review for PRs, git ranges, files, or uncommitted changes

Session-based work execution with progress tracking and knowledge capture

Create implementation plans from requirements

Chart or resequence a multi-unit work horizon — destination, thin claimable streams, order, and NEXT — user-approved map for continue to consume

Agent personality, voice guidance, and communication facts. Two-layer storage — user-space ~/.agent-tools/personify.md (shared defaults) plus optional project .agent-tools/personify.md (local override/delta). Enforces size limits and proactive maintenance. Invocable as /personify for interactive init or collaborative edit+cleanup. Agents load both layers; project AGENTS.md links the project file.

Research-driven audit of product positioning, messaging, competitive stance, and go-to-market readiness

Generate product content — one-pagers, pitch decks, README rewrites, comparison pages — from positioning research and codebase artifacts

Research-driven positioning exercise — competitive research, guided framework application, and positioning document generation

Product strategy skill covering positioning frameworks, competitive research methodology, messaging principles, and go-to-market patterns. Research-first — agents conduct extensive primary web research before analysis.

Discover and author NL test specifications through app scanning, doc import, or guided conversation

Initialize Sentinel QA testing with Playwright Test Agents integration

QA skills for testing workflows — visual inspection tools and NL spec authoring for Playwright Test Agents

Setup (or re-setup) a project for swarm + workflow — author the project charter from detected evidence, link it from AGENTS.md (with conditional CLAUDE.md/GEMINI.md symlinks), and write the .agent-tools/ umbrella with swarm config. Idempotent and never destructive.

Guide projects through Clean Architecture patterns with strict layer boundaries, vertical slicing, and dependency inversion. Use when implementing features following Clean Architecture, creating domain models, application use cases, or validating architectural compliance. Language-agnostic with support for Python, TypeScript, C#, and extensible to others.

Testing strategy for agents. Covers strategy selection (TDD, spec-first, property-based, contract, characterization), Red-Green-Refactor, tracer bullet testing, behavior-driven test design, mocking, property testing, and test quality verification. Language-agnostic with AI-specific anti-patterns.

Discover and refine requirements through guided conversation

Browser automation using Chrome DevTools MCP and the agent-browser CLI. Use when the user needs to navigate websites, fill forms, take screenshots, scrape data, or test web applications.

Explore a single fuzzy idea into a framed concept ready for refinement — collaborative HITL diverge, user-gated converge

Sweep planning/ for fully-completed work, verify completion against git history and PM tooling, present a summary for approval, then purge the confirmed-complete item directories and stale meta files (git-recoverable).

Commit, push, and open a PR

Commit and push

Create a git commit

Family of focused, safe git skills for conventional commits, push + PR flows, and worktree-based parallel development. All skills enforce strict rules around git command usage.

Resume the most recent paused swarm run. Reconciles the saved state.yml against disk + PM ground truth, surfaces any drift for confirmation, then re-enters the orchestration loop. Resilient to clean TERMINAL_PAUSE and to crashes/reboots/network loss.

Create a git worktree and enter it for immediate work

QA inspection tools for visual testing -- video frame extraction, screenshot analysis

Drive or analyze a /swarm test-harness run. Given a scenario name, generate a fresh run and hand off the orchestrator step; given a completed run-dir, check the scenario's hard invariants, judge its observation checklist against the role logs, cluster recurring issues, and propose evidence-linked evolve-style improvements to the swarm role templates and skills. Project-scoped repo-development tool for agent-tools.

Visual design micro-patterns from detail.design for evaluating and improving UI polish — motion, accessibility, typography, and interaction details that make big differences.

Unified project audit — orchestrates domain-specific agents across code, tests, API, frontend, docs, repo infrastructure, and QA coverage into a single deduplicated report

Import and adapt skills written for other agent platforms into the canonical skill corpus. Detects agent-specific patterns, applies safe generalizations, collaborates on markup decisions, and produces clean, portable, well-structured canonical skills.

Fixture family skill exercising agent markup resolution.

Fixture top-level colon-named skill (swarm-test pattern) exercising OpenCode top-level command emission and skills/ removal.

Fixture family skill exercising agent markup resolution.