nix-config

Restate the immediately preceding assistant response in plain, concise human language without jargon while preserving important details and useful formatting. Use when the user says "bro", asks for a simpler explanation, or wants a technical answer made easier to read.

Independently review a candidate software change against its original plan, repository constraints, behavior, tests, design quality, compatibility, security, operations, and verification evidence without modifying files. Use for review-only requests, fresh-context $ship review handoffs, or evidence-led diff assessment. Do not use to implement fixes, refactor code, resolve existing GitHub review threads, perform product or architecture design, or approve a change whose governing plan is missing.

Deploy one release-ready immutable software candidate through the repository's established deployment path, observe declared health signals, and stop, roll back, or hand off forward repair when rollout conditions fail. Use only when the user explicitly invokes $deploy with a target environment and deployment authority. Do not use for implementation, release-readiness review, CI-only verification, implicit production mutation, or a candidate lacking identity, recovery, and health evidence.

Inspect and control live Hunk diff-review sessions through the non-interactive `hunk session` CLI. Use when the user has Hunk running, wants an interactive diff walkthrough, or wants inline Hunk comments. The daemon uses localhost, so sandboxed agents must request network or sandbox access before discovering sessions.

Resolve system boundaries, contracts, data and control flow, quality scenarios, failure behavior, and architectural tradeoffs for an active Neo task. Use only with an architecture-stage handoff generated by $neo. Do not use directly, reduce architecture to technology selection, or implement code.

Resolve system boundaries, contracts, data and control flow, quality scenarios, failure behavior, and architectural tradeoffs for an active Neo task. Use only with an architecture-stage handoff generated by $neo. Do not use directly, reduce architecture to technology selection, or implement code.

Convert approved Neo context and design into thin vertical implementation slices with focused verifiers, compatibility, rollout, rollback, and replan conditions. Use only with a delivery-stage handoff generated by $neo. Do not use directly, create horizontal layer batches, or implement the slices.

Convert approved Neo context and design into thin vertical implementation slices with focused verifiers, compatibility, rollout, rollback, and replan conditions. Use only with a delivery-stage handoff generated by $neo. Do not use directly, create horizontal layer batches, or implement the slices.

Ground an active Neo task in repository evidence, clarify the problem, and separate facts, user statements, inferences, assumptions, and unknowns. Use only with a discover-stage handoff generated by $neo. Do not use directly, design a solution, select architecture, or implement code.

Ground an active Neo task in repository evidence, clarify the problem, and separate facts, user statements, inferences, assumptions, and unknowns. Use only with a discover-stage handoff generated by $neo. Do not use directly, design a solution, select architecture, or implement code.

Investigate and route consequential software-engineering changes through explicit product, architecture, program-design, prototype, and delivery decisions before implementation. Use when the user explicitly invokes $neo for a complex, ambiguous, cross-boundary, persistent-data, security-sensitive, or expensive-to-reverse change. Do not use for implementation, ordinary lightweight planning, code review, or questions about The Matrix.

Consolidate approved Neo artifacts into a versioned implementation brief, process clarification or requested design changes, invalidate affected decisions, and obtain explicit final approval. Use only with a finalize-stage handoff generated by $neo. Do not use directly or begin implementation from a stale or unapproved brief.

Consolidate approved Neo artifacts into a versioned implementation brief, process clarification or requested design changes, invalidate affected decisions, and obtain explicit final approval. Use only with a finalize-stage handoff generated by $neo. Do not use directly or begin implementation from a stale or unapproved brief.

Resolve users, outcomes, scenarios, non-goals, constraints, harms, and product-prototype decisions for an active Neo task. Use only with a product-stage handoff generated by $neo. Do not use directly, choose software architecture, or implement code.

Resolve users, outcomes, scenarios, non-goals, constraints, harms, and product-prototype decisions for an active Neo task. Use only with a product-stage handoff generated by $neo. Do not use directly, choose software architecture, or implement code.

Shape modules, interfaces, types, invariants, ownership, data structures, dependency direction, and main or failure call paths for an active Neo task. Use only with a program-stage handoff generated by $neo. Do not use directly, emit production code, or substitute a file inventory for program design.

Shape modules, interfaces, types, invariants, ownership, data structures, dependency direction, and main or failure call paths for an active Neo task. Use only with a program-stage handoff generated by $neo. Do not use directly, emit production code, or substitute a file inventory for program design.

Investigate and route consequential software-engineering changes through explicit product, architecture, program-design, prototype, and delivery decisions before implementation. Use when the user explicitly invokes $neo for a complex, ambiguous, cross-boundary, persistent-data, security-sensitive, or expensive-to-reverse change. Do not use for implementation, ordinary lightweight planning, code review, or questions about The Matrix.

Manage an implementation-ready software plan through repository orientation, disciplined TDD, verification, fresh-context independent review, bounded remediation, optional existing-PR feedback handling, and a release-ready decision. Use only when requirements and consequential design decisions are already settled and the user explicitly invokes $ship or supplies a validated ship handoff. Stop for missing plan information or replanning; do not perform research, product or architecture design, deployment, release, publishing, or merge.

Design, create, revise, evaluate, instrument, secure, and release reusable agent skills from observed successful workflows or recurring reviewed failures. Use when deciding whether behavior belongs in a skill, retrieval source, script, tool, hook, prompt, or repository instruction; creating or updating a skill package; defining triggers and progressive disclosure; setting up skill telemetry and tracing; selecting model and reasoning-effort lanes; comparing a candidate with no-skill or built-in skill creators; promoting reviewed traces into regression cases; or making an evidence-backed release decision. Do not use for one-off task instructions, speculative workflows with no inspected examples, or merely installing an existing skill.

Implement or repair predictable software behavior through disciplined test-driven development: a living behavioral test list, one meaningful failing test at a time, the smallest real passing change, optional green-state refactoring, and explicit evidence. Use for behavior changes, regression fixes, interface-driving examples, or a validated $ship behavior handoff. Do not use for pure refactoring, unsettled requirements or architecture, deployment, research, or behavior without a timely trustworthy executable oracle.

Restate the immediately preceding assistant response in plain, concise human language without jargon while preserving important details and useful formatting. Use when the user says "bro", asks for a simpler explanation, or wants a technical answer made easier to read.

Inspect and control live Hunk diff-review sessions through the non-interactive `hunk session` CLI. Use when the user has Hunk running, wants an interactive diff walkthrough, or wants inline Hunk comments. The daemon uses localhost, so sandboxed agents must request network or sandbox access before discovering sessions.

Review and refactor existing code for safer change, clearer design, lower complexity, and long-term maintainability. Use when asked to review code health or design, simplify or restructure code, remove technical debt or code smells, improve abstractions or module boundaries, apply or assess design patterns, prepare code for a feature, or make an existing codebase easier for humans to understand without unintentionally changing behavior. Also use when a requested cleanup overlaps tests, public APIs, compatibility, or migration risk.

Review and refactor existing code for safer change, clearer design, lower complexity, and long-term maintainability. Use when asked to review code health or design, simplify or restructure code, remove technical debt or code smells, improve abstractions or module boundaries, apply or assess design patterns, prepare code for a feature, or make an existing codebase easier for humans to understand without unintentionally changing behavior. Also use when a requested cleanup overlaps tests, public APIs, compatibility, or migration risk.

Investigate a question against high-trust primary sources and report evidence-backed findings. Use when the user wants a topic researched, documentation or API facts gathered, or reading legwork delegated. Return findings in chat by default; create a Markdown artifact only when the user requests persistent output or applicable instructions require it.

Investigate a question against high-trust primary sources and report evidence-backed findings. Use when the user wants a topic researched, documentation or API facts gathered, or reading legwork delegated. Return findings in chat by default; create a Markdown artifact only when the user requests persistent output or applicable instructions require it.

Triage and address GitHub pull-request review threads through verified fixes. Use when the user asks to inspect, address, respond to, or resolve PR review feedback. Support local-only remediation and explicit end-to-end mode; push, reply, or resolve remote threads only when the user authorizes PR updates.

Triage and address GitHub pull-request review threads through verified fixes. Use when the user asks to inspect, address, respond to, or resolve PR review feedback. Support local-only remediation and explicit end-to-end mode; push, reply, or resolve remote threads only when the user authorizes PR updates.