Coding standards, frameworks, refactoring, and code review (18.4K skills)
Plan how a single canonical agent behavior contract can run across tools such as Codex, Claude Code, Gemini CLI, and Cursor. Get a runtime matrix covering entry points, commands, adapters, memory access, limitations, and verification steps.
Create thin adapters for Codex, Claude Code, Gemini CLI, Cursor, and other listed runtimes from one canonical AGENTS.md core. This helps generated repositories map shared instructions to each tool’s local conventions without duplicating the core.
Checks whether an Agentlas agent repository has the files and validation steps needed for public distribution, including GitHub release, Codex plugin submission, Claude adapters, and one-line installation. It also checks that public packages exclude secrets and private materials.
Builds a single Agentlas agent, a coordinated team, or a package for an existing agent. It clarifies the work through an interview, researches relevant tools and examples, then creates instructions and supporting package files in English.
Prepare a project folder for Agentlas so future agent work can use project memory and sitemap data. It guides safe activation, adds missing state files without replacing existing content, and routes durable memory updates through a review process.
Plan how a shared agent core can run across tools such as Codex, Claude Code, Antigravity, and Cursor. Define each runtime’s entry point, commands, adapters, memory access, limitations, and verification method in a comparison matrix.
Design an agent team around the user’s job and target runtime. Define when separate roles are useful, how work passes between them, and who owns memory, policy, and evaluation; receive a role list and team topology.
Convert or repair an existing agent or team package for Agentlas. It preserves useful behavior, adds routing, memory, evaluation, and runtime contracts, then verifies package structure and public safety before a handoff or release.
Build, repair, or package Agentlas agents and teams with Hephaestus. The skill delegates the request to the shared hep-build instructions, which define the workflow for producing or fixing an agent or team package.
Adds and audits lifecycle records for Agentlas skills, including a candidate registry, trial evidence, and Curator decisions. It checks promotion gates, approval status, evidence gaps, and rollback readiness before a skill can become first-class.
Prepare an agent package for public distribution by checking required Codex and Claude Code files, install commands, verification scripts, and release safety boundaries. Useful when publishing a plugin or open-source package.
Use this guide to create or repair `.agentlas/routing-card.json`. It explains what each field should contain, which fields affect agent matching, and how misplaced sentences or descriptions can prevent a card from being found.
Creates lightweight runtime adapters for a canonical agent core, so one set of instructions can be mapped into Codex, Claude Code, Antigravity, Cursor, and generic AGENTS.md conventions. Each adapter follows its runtime's local conventions.
Compare installed Copilot session event schemas with the Session Log canvas baseline, classify changes, and guide code, fixture, documentation, and baseline updates. It helps keep new session data visible while preserving readability for sessions recorded by earlier versions.
Integrate Git branches with merge, rebase, or rebase onto a new base. The skill guides conflict resolution, can pause for review after fixes, and reports the outcome. It pushes only when that action is explicitly authorized.
Checks Git settings with a read-only baseline, identifies missing identity, editor, diff, or merge tool values, and proposes minimal fixes. It applies only the groups the user confirms, preserves existing custom settings, and handles signing or safe-directory help when requested or relevant.
Create one local Conventional Commit from paths the user selects and confirms, or generate a ready-to-paste message for already staged changes. The message-only mode leaves the Git index untouched, and neither mode pushes changes remotely.
Resolve which cspell configuration a project command actually loads, then curate project terms and ignored paths. The workflow updates the active config or dictionary, reruns the same spell check, and reports the changes and remaining issues.
Review an implementation by comparing its plan, change record, research, and validation evidence. Record scoped findings with severity, supporting evidence, resolution conditions, and suggested routes for accepted follow-up work.
Use this skill to carry out an approved implementation plan in its declared scope. It tracks completed tasks, updates the plan when new information appears, and records behavior changes, validation results, blockers, and remaining work for a clear handoff.
Create or update an evidence-based implementation plan that organizes work into phases and tasks, records requirements, risks, dependencies, and open decisions, and can include an independent critique to help judge readiness.
Get a guided, step-by-step explanation of code, interfaces, features, prompts, or tracking documents. The walkthrough links to evidence, reviews the target before explaining it, and records material decisions or requested changes when they arise.
Prepare pull requests and code reviews by collecting commit history and unified diffs between branches into structured XML. Filter files by extension or path, then list changed files or inspect large diffs in manageable chunks.
Check an implementation plan against supplied requirements and evidence without changing the plan. Receive one consolidated assessment with actionable findings, severity, evidence, and the smallest useful corrections; follow-up reviews also track whether earlier findings were resolved.