Create ShinkaEvolve task scaffolds from a target directory and task description, producing `evaluate.py` and `initial.<ext>` (multi-language). Use when asked to set up new ShinkaEvolve tasks, evaluation harnesses, or baseline programs for ShinkaEvolve.

Run existing ShinkaEvolve tasks with the `shinka_run` CLI from a task directory (`evaluate.py` + `initial.<ext>`). Use when an agent needs to launch async evolution runs quickly with required `--results_dir`, generation count, and strict namespaced keyword overrides.

Load top-performing Shinka programs into agent context using `shinka.utils.load_programs_to_df`, and emit a compact Markdown bundle for iteration planning.

Convert an existing codebase in the current working directory into a ShinkaEvolve task directory by snapshotting the relevant code, adding evolve blocks, and generating `evaluate.py` plus Shinka runner/config files. Use when the user wants to optimize existing code with Shinka instead of creating a brand-new task from a natural-language description.

Iterative development with AI review. Provides RLCR (Ralph-Loop with Codex Review) for implementation planning and code review loops.

Refine an annotated implementation plan into a comment-free plan and a QA ledger while preserving the gen-plan schema.

Generate a structured implementation plan from a draft document. Validates input, checks relevance, analyzes for issues, and generates a complete plan.md with acceptance criteria.

Consult Gemini as an independent expert with deep web research. Sends a question or task to Gemini CLI and returns a research-backed response.

Session wrap-up: commit pending changes, push, write structured state for the next session to resume from. Eliminates 'glue sessions' that exist only to commit/push, and preserves debug context across sessions.

Agent Loop for pixel-perfect UI adjustments: extract a structured spec from the user's intent, apply changes one property at a time, self-verify via screenshot, and converge instead of ping-ponging. Designed for high-fidelity UI cloning and spacing/style iteration.

TRIGGER when writing, modifying, or reviewing any Rust code that involves tracing, logging, spans, events,

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes

Perform a strict, evidence-based review of the current branch or working-tree changes without modifying files. Use when the user asks for a strict review, deep review, senior review, merge assessment, or explicitly invokes `$review-strict`.

Agent Loop for tasks that need research before implementation: check existing research first, run structured investigation (bb-browser/context7/code study), produce a Brief, then implement against the Brief. Prevents repeated research on already-studied topics and research-to-implementation drift.

Refresh fact-based sections of crates/AGENTS.md (STRUCTURE, WHERE TO LOOK, COMPLEXITY HOTSPOTS, KNOWN BROKEN) from current codebase state. Use after refactoring, renaming files, adding crates, or when the agent notices AGENTS.md info is stale.

Agent Loop that runs local pre-flight CI checks, auto-fixes issues, creates/pushes the PR, monitors CI, and loops until all checks pass. Replaces the manual chain of preflight → create-pr → babysit-pr.

Run an independent adversarial review loop for a Markdown implementation plan. Use when the user asks to review, validate, stress-test, or challenge a plan before implementation, including requests for a second opinion or a plan review loop.

Record a lesson learned into crates/AGENTS.md (CONVENTIONS or ANTI-PATTERNS section). Use after discovering a project-specific gotcha, resolving a tricky bug, or when the user says 'remember this' about a coding pattern. Append one concise rule that will prevent the same mistake in future sessions.

Drive the UniClipboard iOS app in a simulator and read its OSLog yourself to diagnose a mobile-sync bug, instead of asking the user to paste logs. Use when debugging why the iOS sync engine or M5 reducer did something, reproducing an iOS sync / clipboard bug on the simulator, or whenever you'd otherwise ask the user "what do the logs say". All output is Swift OSLog under subsystem `app.uniclipboard` — the Rust core (uc-mobile) emits no logs of its own; the Swift shell logs the reducer's decisions.

Generate bilingual, user-facing release changelogs from Git history. Use when the user asks to generate a changelog, prepare release notes, or update `docs/changelog` for a release and provides a base tag or commit.