cortex-command

Structured feature development lifecycle with phases for research, specification, planning, implementation, review, and completion. Use when user says "/cortex-core:lifecycle", "start a lifecycle", "lifecycle research/specify/plan/implement/review/complete", "start a feature lifecycle", or wants to build a non-trivial feature with structured phases.

Development entry point that analyzes requests and routes to the appropriate workflow. Use when user says "/cortex-core:dev", "what should I work on", "start working on", "dev hub", "where do I start", "next task", "what's next", or describes a feature without naming a specific skill.

Prepare a backlog item for execution by running it through Clarify → Research → Spec. Use when user says "/cortex-core:refine", "refine backlog item", "prepare for overnight", or "prepare feature for execution". Produces cortex/lifecycle/{slug}/research.md and cortex/lifecycle/{slug}/spec.md, then sets status:refined on the backlog item.

Parallel research orchestrator. Use when the user says "/cortex-core:research", "research this topic", "investigate this feature", or when /cortex-core:refine delegates its research phase. Dispatches 3–10 parallel agents across independent angles, synthesizes into research.md or conversation output.

Create git commits with consistent, well-formatted messages. Use when user says "commit", "/cortex-core:commit", "make a commit", "commit these changes", or asks to save/checkpoint their work as a git commit.

Guide the user through the morning report after an overnight session. Walks Executive Summary and report sections, collects answers to deferred questions, advances completed lifecycles to Complete, and auto-closes backlog tickets. Use when the user says "/morning-review", "morning review", or "what happened overnight".

Parallel adversarial review — dispatches reviewer agents on distinct challenge angles, then synthesizes findings with an Opus agent. Use when user says "critical review", "pressure test", "adversarial review", or "challenge from multiple angles". Auto-triggers in the lifecycle for Complex + medium/high/critical features before spec approval.

Ideation research for topics not ready for implementation — checks aim, investigates the problem space, then decomposes findings into backlog tickets grouped by epic. Use when user says "/cortex-core:discovery", "discover this", "break this down into tickets", "decompose into backlog", or wants to understand a topic before committing to build. Requires a topic argument; for "what should I work on" or "next task" routing without a specific topic, use /cortex-core:dev instead.

Plan and launch autonomous overnight development sessions. Selects eligible backlog features, presents a session plan for approval, and hands off to the runner for unattended execution. Use when user says "/overnight", "start overnight session", "launch overnight", or wants to run features autonomously overnight.

Manage project backlog items as individual markdown files with YAML frontmatter. Use when user says "/cortex-backlog:backlog", "backlog add", "backlog list", "backlog pick", "add to backlog", "show backlog", "archive backlog item", "what's ready", "pick a backlog item", or asks to create/view/manage/select items.

Systematic 4-phase debugging for skills, hooks, lifecycle, and overnight runner issues. Use when: "fix this bug", "why is this failing", "make this test pass", "skill not working", "hook not running", "diagnose this", or unexpected agentic-layer behavior. Finds root cause, fixes, and verifies.

Interview-only sub-skill that produces a structured Q&A markdown block for /requirements-write to synthesize. Adopts mattpocock interview patterns. Invoked by the /cortex-core:requirements orchestrator.

Use /cortex-core:requirements to gather requirements or define project scope. Thin orchestrator that routes to /requirements-gather then /requirements-write; disable-model-invocation:true — invoked only by explicit slash command.

Synthesize-only sub-skill that turns a /requirements-gather Q&A block into a v2-compliant cortex/requirements/{project|area}.md; never conducts an interview itself. Invoked by the /cortex-core:requirements orchestrator.

Compose structured backlog ticket bodies using the Why/Role/Integration/Edges/Touch-points template. Use when user says "ticket authoring", "compose", "write a ticket body", or "author a backlog item". Two subcommands: interview (guided prompts) and compose (autonomous, context-driven).

General-purpose priming interview — a structured "grilling" loop to help you think through a topic, then a concise brief. A thinking-partner interview, NOT backlog-ticket authoring; for a ticket body use /cortex-core:backlog-author.

Create GitHub pull requests with well-crafted titles and descriptions. Use when user says "pr", "/cortex-core:pr", "create a pr", "open a pull request", "submit a pr", "make a pull request", or asks to get their changes ready for review.

Reference for writing and editing skills well — the vocabulary and principles that make a skill predictable. Reach for it while authoring or pruning cortex-command skills.

Review a GitHub pull request with a single full-context reviewer and present a deterministic verdict. Use when user says "/pr-review", "/pr-review <number>", "review this PR", "review PR 123", "review my pull request", or asks for a code review of an open pull request.

Orchestrates Android development tasks including project creation, deployment, SDK management, and environment diagnostics using the `android` command-line tool.

Analyzes Android build files and R8 keep rules to identify redundancies, broad package-wide rules, and rules that subsume library consumer keep rules. Use when developers want to optimize their app's size, remove redundant or overly broad keep rules, or troubleshoot Proguard configurations.

Invalid unknown-subcommand-flag fixture — bogus flag on a known subcommand must produce E102

Valid subcommand-flag fixture — known subcommand with all required flags must produce zero violations

Path-prefixed true-positive fixture — cortex-worktree-create with a bin/ prefix followed by flags without --feature must flag E101

Env-prefix true-positive fixture — cortex-worktree-create preceded by an env assignment without --feature must flag E101

Baseline true-positive fixture — bare invocation of cortex-worktree-create without --feature must flag E101

Piped true-positive fixture — cortex-worktree-create after a pipe separator without --feature must flag E101

Positional-tail true-positive fixture — cortex-worktree-create with a path-like positional arg but no --feature must flag E101

Mixed-span fixture — one fenced span holds both a path-mention (dropped) and a real invocation missing --feature (kept)

command -v probe fixture for contract lint — fenced probe with redirection tail

which/type/hash probe fixture for contract lint — probe operands are not invocations

Hook filename fenced-block mention fixture for contract lint

Hook filename inline-code mention fixture for contract lint

Hook filename table-cell mention fixture for contract lint

Demonstrates E104 — not_argparse binary without ledger entry

Ledger-suppresses fixture for contract lint

Non-argparse-exempt fixture for contract lint

Prose-mention-only fixture for contract lint

Sentinel-ignored fixture for contract lint

Missing required flag fixture for contract lint