tylerakins/ai-agent-skills

Collect and verify GitHub PR feedback, repair supported issues within authorized scope, and draft concise thread replies. Use when asked to address PR comments or review feedback. Get approval for the exact replies and their intended thread resolutions together before publishing either action.

Investigate coding requirements and turn them into an ordered, verifiable implementation plan. Use when asked to plan a feature, ticket, or significant change, assess an approach, or break work into tasks. Does not implement the plan.

Draft and check concise technical prose for chat, reviews, replies, tickets, PRs, commits, documentation, and code comments. Use when writing or revising human-facing text; preserve destination contracts and publication approval.

Design and evolve API and module contracts with explicit behavior, validation, error semantics, authorization, and compatibility. Use for endpoints, schemas, service boundaries, SDKs, or shared interfaces. Follow existing protocol and repository conventions rather than imposing a new API style.

Design, modify, debug, and verify continuous integration and deployment workflows. Use for build/test pipelines, release automation, deployment gates, and CI reliability or performance. Preserve trust boundaries and existing required checks; pipeline changes do not authorize deployments or repository policy changes.

Review local diffs, branches, implementation plans, or GitHub pull requests against requirements and repository evidence. Use for requested code or plan reviews and delegated independent review. Draft concise findings; publish GitHub reviews only after approval of the exact verdict, text, and locations.

Clarify goals and challenge consequential assumptions through focused questions and evidence. Use when asked to interview, question, or stress-test an idea or plan, or when unresolved intent would materially change the approach. Stop when the important decisions are clear; respect delegated judgment and existing authorization.

Explain concepts, code, documents, or decisions in simple language with a short TL;DR, concrete examples, and depth matched to the reader; use for plain-English explanations, walkthroughs, or summaries rather than implementation or prose editing.

Build and improve user interfaces with coherent visual design, responsive layout, accessible interaction, and reliable state. Use for pages, components, forms, frontend behavior, and UI polish. Follow the project's framework and design system and verify rendered behavior rather than relying on source or screenshots alone.

Prepare scoped Git commits and concise GitHub PR titles and descriptions from the actual diff and verification. Use when asked to stage, commit, push, draft or update a PR description, or open a PR. Perform only the authorized actions.

Execute an accepted implementation plan through scoped changes, meaningful verification, independent final review, and repairs. Use when asked to build or implement a plan, including a Plan-to-Build handoff. Preserve requirements and authorization; commit, push, and publication remain separate actions.

Read Jira issues into sourced requirements and unresolved decisions, or draft, create, and edit concise Jira descriptions and status comments. Use when a Jira key or URL is supplied for work, or Jira writing is requested. Preserve explicit ADF formatting for interactive acceptance criteria, code marks, and mentions.

Design, rewrite, or review a project's README as its public entrypoint. Use when readers need to understand the project, choose a path, get started, and find deeper documentation. Use technical-documentation for source-derived guides and references.

Assess a specific release, prepare staged rollout and recovery, and verify production outcomes. Use when asked whether a change is ready to ship, to prepare a launch, or to conduct an authorized rollout. Readiness assessment does not authorize deployment, feature activation, rollback, or external announcements.

Predict what a named reviewer may flag from their feedback in a specific repository. Use for reviewer-specific pre-review or preference analysis. Separate historical preferences from verified defects; do not replace correctness review or imply the reviewer will approve.

Diagnose bugs, failing tests, crashes, wrong output, flaky behavior, and build or production errors using reproduction, competing hypotheses, and causal evidence. Use when asked to investigate or fix unexpected behavior. Repair supported causes within authorization and verify regression coverage rather than masking symptoms.

Plan and implement compatible transitions for APIs, services, dependencies, configuration, and persisted data, including deprecation and removal. Use when replacing an interface or system, retiring behavior, or changing schemas with active consumers. Verify staged compatibility, cutover, and recovery within authorization.

Design, author, evaluate, or package agent skills using the destination's conventions and validation requirements. Use when creating or revising a skill, checking its activation triggers, or preparing a skill package.

Create, update, or review technical documentation from repository sources and primary references, including developer and user guides, API or factual references, onboarding material, technical specifications, release notes, and architecture decision records. Use readme-design for project READMEs and explain-it for one-time explanations.

Write meaningful tests for requested behavior, boundaries, failures, and state or tenant isolation. Use when adding coverage, reproducing a bug, or writing regression tests. Follow the repository's runner and conventions; do not expand test-authoring requests into unrequested production changes.

Route ambiguous or multi-workflow engineering requests to the right ai-agent-skills procedures. Use when a task spans several workflows or no single skill clearly applies. For a clear single-skill request, load that skill directly instead.

Write or simplify code with clear control flow, focused responsibilities, and only the abstraction current requirements justify. Use when implementing a scoped change, reducing unnecessary complexity, or refactoring for readability. Preserve repository conventions and contracts; avoid unrelated cleanup.

Write, review, and optimize SQL queries, procedures, and application SQL for SQL Server, MySQL/MariaDB, PostgreSQL, BigQuery, Snowflake, Databricks SQL, Oracle, Vertica, and Teradata; preserve result semantics and verify tuning against plans and measured workloads.