savenkovigor/railing
Engineering guardrails that keep AI coding agents on track.
You should use this skill when checking or validating files against best-practice references, including code, documentation, configuration, and skill files.
Manual-only debug command. Do NOT trigger this automatically - only invoke it when the user explicitly runs it by name. It snapshots the AI context init state (instructions, commands, skills, MCP servers, and tools loaded into the session before the user typed anything) for debugging purposes.
You should use this skill when: asked to help make decisions based on the conversation so far, asked what is left to decide before work starts, a discussion ends with several open options and no chosen one
You should use this skill when: designing or auditing a design system; extracting design tokens from code; validating design tokens and UI consistency; reviewing implementation against spec; ensuring design best practices.
You should use this skill when a team asks which engineering practices are missing and in what order to adopt them (maturity assessment, practice gap analysis, "what should we improve first?"). Do not use it for bug fixes, code review, or architecture decisions.
You should use this skill when: a language, framework, or SDK version was upgraded, an API changed or was deprecated, old usage must be refactored across a codebase, or migration work should be split into focused skills.
You should use this skill when: asked to build or update a glossary, dictionary, or ubiquitous language of a project, asked to collect the terms a team uses from its issue tracker, docs, and code, asked which word the team uses for a domain concept
You should use this skill when: a task is completed, a session is long: its context was compacted, it made dozens of tool calls, the agent started repeating earlier steps, or more than half of the context window is used when a token counter is visible, or when a scheduled or manual review of recent agent sessions is requested, to reflect on the sessions, propose the 3 most important improvements to AI context files that would improve speed or quality next time, and record the run in a journal of reviewed sessions and decisions
You should use this skill when: asked to clean up, review, or check a report in notes/, asked to verify that a technical markdown document matches the code
You should use this skill when: asked to write a technical report about existing code into notes/<task_id>/, asked for an architecture overview of classes, modules, or a subsystem, asked to explain data flow, ownership, or lifecycle before working on a feature
You should use this skill when: asked to review a file, uncommitted changes, or the current branch, asked whether a change is safe to merge, asked to check that a change does what its description says
You should use this skill when: breaking down a plan into tasks, converting a discussion into actionable issues, splitting a proposal into tickets, creating issues from meeting notes or design decisions. Works with any issue tracker - infers GitHub Issues, Jira, Linear, GitLab, etc. from project context.
You should use this skill when: asked to explain, map, or onboard onto an unfamiliar codebase or repository, asked how a project is structured or where to change something in it, asked to drill into one area of a codebase after a first overview