aili-workflows

Create, update, or validate a project-local AGENTS.md from the shared AILI template when initializing or checking repository agent rules.

Run the AILI delivery lifecycle for /ideate, /define, /build, and /ship; use for idea shaping, spec/test definition, autonomous BUILD package queues, review-repair closeout, or backend routing without exposing internal stage commands.

Analyze explicit user-provided or approved evidence such as sanitized OpenCode session exports, selected transcripts, git history, implementation notes, or task records to produce a safe report-first workflow retrospective; do not claim global history access, commit raw sessions/logs, or directly edit protected harness surfaces.

Diagnose user-reported AILI/ROSE harness or workflow behavior problems without editing files; use when the user says a command, skill trigger, subagent packet, memory flow, installer, docs, or agent prompt behavior is wrong and wants to know where the issue lives and how to fix it. Do not use for normal product-code bugs; hand off approved harness modifications to harness-evolution.

Use for `/local-review` or lifecycle-triggered local review gates over worktree diffs, refs, PRs, or OpenSpec changes; owns target resolution, evidence scope, categorized review reports, verdicts, explicit repair/re-review, and no-remote-mutation boundaries without overriding OpenCode `/review`.

Use when ROSE needs context-saving subagent dispatch: single read-only scouting for noisy repository evidence, or splitting two or more independent investigation, implementation, review, testing, documentation, or security work packages across subagents without shared mutable state, overlapping edits, or sequential dependencies.

Canonical AILI requirements-grilling workflow for evidence-first requirements clarification, change-interviewer compatibility, Chinese interview packet generation, OpenSpec interview.md write-back readiness, domain-modeling, ADR gating, and grill/interview packet trigger phrases before BUILD.

Orchestrates post-implementation review before final PASS. Use after non-trivial changes to fan out only relevant code, test, security, coverage, PR-test, AI-regression, silent-failure, convergence, browser-QA, E2E-artifact, agent-output, or OSS-sanitizer lanes, reconcile findings, run a bounded fix loop, and gate completion without pushing or merging.

Use the ROSE project-local SQLite memory system through the bundled memory_cli.py. Use this for task checkpoints, requirement memory, durable findings, focused retrieval packs, and writeback receipts.

Create, revise, and validate Agent Skills for this repository. Use when the user wants to add a new SKILL.md, improve an existing skill, convert repeated workflow instructions into a skill, optimize skill triggering, split references/scripts/assets, or check whether a skill is too broad, too long, or likely to over-trigger.

Discovers and invokes agent skills. Use when starting a session or when you need to discover which skill applies to the current task. This is the meta-skill that governs how all other skills are discovered and invoked.

Research mature public project patterns before方案/implementation for IDEATE, DEFINE, BUILD planning, or ordinary chat prior-art needs such as "看看别人怎么做", "GitHub 上别人怎么做", "look at how others do it", "reference mature projects", "how do mature projects handle this?", "compare prior art", industry/GitHub similar projects, or explicit user-requested research; wraps read-only web-researcher evidence and does not trigger for GitHub issue/PR triage, local code-only review, implementation, or command creation.

在非平凡规划、方案、编辑、审查或完成声明前，先用仓库证据锚点确认项目事实；当本地约定、peer pattern、测试、配置或现有实现会影响方案/implementation 时必须先取证；无证据的项目判断必须标为 Hypothesis、Open Question、Unverified、委托调查或 blocked。

Grounds every implementation decision in official/API documentation. Use when you want authoritative, source-cited code free from outdated patterns; when official/API docs such as DeepSeek, SDK/provider references, framework docs, changelog-sensitive behavior, fast-changing/version-sensitive sources, model uncertainty, or explicit user-requested verification can materially affect a方案 before implementation.

Generate a repository-local, evidence-grounded Markdown test document from specs, proposals, evidence-backed方案, plans, issues, or user descriptions; ask only for missing product decisions, map requirements/research risks/UNVERIFIED planning gaps to test cases, and persist the test plan before summarizing.

Repair build, typecheck, lint, or test failures with a root-cause-first minimal workflow; route read-only investigation before approved implementation, do not change dependencies or lockfiles without approval, and never bypass or skip failing gates to claim success.

Code-review quality gate and test-enhancement rubric. Use when ROSE needs sharper review output contracts, severity and risk prioritization, evidence anchors, negative review cases, fixture or golden-output drift guidance, or a Chinese review report profile; do not use as a duplicate code-review agent or final PASS authority.

Review comment, JSDoc, TODO, README, and docs-to-code factual consistency, stale or low-value comments, and Chinese comment/variable appropriateness; do not use for general code review, broad documentation rewrites, or auto-editing comments without approval.

Report-first audit for AILI/ROSE harness routing, context cost, review fan-out, subagent parallelism, trigger noise, false PASS risk, and evidence-loss risk. Use when optimizing harness quality or cost; do not edit core harness controls and route approved changes to harness-evolution.

Run a non-destructive open-source, npm, or public release readiness checklist covering package metadata, provenance, license, internal artifact exposure, and dry-run evidence; complements security, shipping, and sanitizer reviews, and must not publish, delete, rewrite history, or mutate release state.

Stress-tests a proposed strategy, interview packet, spec, plan, task breakdown, subagent reconciliation, review, or completion claim before acceptance; use after a first draft exists, before write-back, before implementation, during reconciliation, or before claiming complete/fixed/verified/ready.

Use before any agent claims a task is complete, fixed, passing, verified, ready, or accepted, especially after implementation, debugging, testing, review follow-up, configuration, or release work.

Review a single academic paper, preprint, DOI, arXiv link, or user-provided PDF/text with source-grounded critique. Use for paper summaries, methodology review, novelty checks, reproducibility concerns, or "review this paper" requests; do not use for multi-paper surveys, systematic literature reviews, citation management, or implementation from a paper.

AI regression scouting routing. Use when agents, prompts, skills, model/tool routing, harness fixtures, or generated-output expectations change and need regression scenarios; do not use for ordinary product-code regressions.

Android native Kotlin/Compose app development, Material 3 UI, accessibility, and Gradle builds.

Guides stable API and interface design. Use when designing APIs, module boundaries, or any public interface. Use when creating REST or GraphQL endpoints, defining type contracts between modules, or establishing boundaries between frontend and backend.

Browser QA routing. Use for local browser-rendered UI verification, DOM/accessibility/console/network checks, screenshots, and Playwright browser evidence; do not use for backend-only changes or production-mutating flows.

Tests browser apps with OpenCode Playwright tools; inspect DOM, console, network, screenshots, and runtime behavior.

Design, critique, or specify honest charts and data visualizations from known data and audience needs. Use for chart selection, misleading-chart audits, accessibility, dashboard visuals, or "make this data easier to understand"; do not use for raw data analysis, PPT deck generation, spreadsheet surgery, or fabricated chart data.

Automates CI/CD pipeline setup. Use when setting up or modifying build and deployment pipelines. Use when you need to automate quality gates, configure test runners in CI, or establish deployment strategies.

Conducts multi-axis code review. Use before merging any change. Use when reviewing code written by yourself, another agent, or a human. Use when you need to assess code quality across multiple dimensions before it enters the main branch.

Simplifies code for clarity. Use when refactoring code for clarity without changing behavior. Use when code works but is harder to read, maintain, or extend than it should be. Use when reviewing code that has accumulated unnecessary complexity.

Structure source-grounded consulting-style analysis for strategy, market, product, operations, or decision memos. Use for MECE problem framing, hypotheses, options, tradeoffs, executive recommendations, or "consulting analysis" requests; do not use for legal/financial advice, implementation, data computation, chart rendering, or unsourced business claims.

Optimizes agent context setup. Use when starting a new session, when agent output quality degrades, when switching between tasks, or when you need to configure rules files and context for a project.

Coverage review and verification sufficiency routing. Use when test coverage, uncovered paths, coverage reports, or verification adequacy need a read-only QA review; do not use for writing tests or running browser/E2E artifacts.

Analyze source data, tables, CSV/XLSX extracts, metrics, or experiment results with explicit data-authenticity checks. Use for profiling, cleaning assumptions, statistical summaries, segmentation, trend analysis, or "analyze this data"; do not use for spreadsheet file surgery, chart design only, database migrations, or claims without data provenance.

Guides systematic root-cause debugging. Use when tests fail, builds break, behavior doesn't match expectations, or you encounter any unexpected error. Use when you need a systematic approach to finding and fixing the root cause rather than guessing.

Manages deprecation and migration. Use when removing old systems, APIs, or features. Use when migrating users from one implementation to another. Use when deciding whether to maintain or sunset existing code.

Records decisions and documentation. Use when making architectural decisions, changing public APIs, shipping features, or when you need to record context that future engineers and agents will need to understand the codebase.

E2E artifact handling routing. Use when traces, videos, screenshots, reports, or failure bundles need controlled repository-local placement; do not use for simple unit tests or production-mutating E2E flows.