Software Development Skills

Ensures engineering work aligns with Architecture Decision Records and catches architectural drift. Use when planning, implementing, reviewing, or when no ADRs exist and bootstrap is needed. Emits ADRCOMPLIANCE. Never silently invents architecture or auto-modifies ADRs.

Guards OpenAPI/proto/GraphQL contracts against breaking changes and drift from implementation. Use on API PRs and releases. Emits CONTRACTREPORT. Never rubber-stamps breaking changes without a versioning/migration plan.

Analyzes impact of proposed changes on existing systems (brownfield projects) with delta spec validation.

Investigates bugs with evidence and analyzes root causes. Use for defects, incidents, flakes, and “why did this break?” work. Modes: investigate (reproduce + safest fix options) and rca (causal chain + prevention, no code). Replaces debugging-investigator and root-cause-analyzer. Never guesses or patches symptoms without evidence.

Turns validated specs into an executable delivery plan: vertical-slice milestones, risks, rollback, then an atomic task graph with AC IDs. Use before coding for features and multi-step work. Replaces implementation-planner and task-decomposer. Never writes production code or redesigns architecture.

Senior mentoring for tradeoffs, knowledge gaps, and growth without taking over execution. Use when the user wants guidance, options, or learning. Modes Mentor (default) vs Executor handoff. Never shames, never silently becomes quality-gate or implementer, and always ends with one concrete next action.

AI Engineering Operating System conductor. Classifies intent, selects a path (lite, standard, incident, architecture, review-only), invokes only the needed skills, threads artifacts, and enforces stop conditions. Use for end-to-end engineering work, “run the OS”, ship a feature, or handle an incident. Never re-implements specialist logic and never runs every skill blindly.

Interviews you to configure project engineering guardrails (code, architecture, security, testing, git/PR, ops, AI agent behavior) and writes them to a durable project file. Use for greenfield setup, “set our rules”, or refreshing standards. Emits GUARDRAILSCONFIG. Never invents team policy as fact, never overwrites an existing guardrails file without confirmation, and never starts feature work.

Plans safe schema/data migrations with expand/contract, backfill, and rollback. Use for DB/search/queue shape changes. Emits MIGRATIONPLAN. Never recommends destructive one-shot prod migrations without a rollback path.

Writes accurate PR descriptions from git diff and commits only—never invented intent or test results. Use when opening or updating a pull request. Emits PRBODY. Prefer the repo’s PR template when present.

Identifies safe, high-value refactors that preserve externally observable behavior. Use for debt, smells, or “clean this up” asks. Never changes behavior silently or recommends large rewrites without characterization tests and incremental plan.

Adversarially clarifies product and engineering requirements before Spec Validator. Use when asks are vague or high-stakes. Emits GRILLNOTES. Never invents product decisions as facts or starts implementation.

Runs red/green/refactor loops against acceptance criteria and TESTSTRATEGY. Use when practicing TDD or when tests should lead implementation. Emits TDDSESSION. Never skips red, never refactors on red, never claims green without command evidence.

Designs a risk-based test strategy mapped to acceptance criteria without writing test code unless asked. Use after planning and before or during implementation. Emits TESTSTRATEGY. Never chases 100% coverage or tests implementation details by default.