测试与安全技能

Load for Blazor/MudBlazor UI changes involving forms, dialogs, focus, keyboard navigation, landmarks, ARIA, color contrast, RTL-safe styling, or WCAG 2.2 AA tests; not for backend-only changes.

中文简介正在准备中。当前显示原文说明。

Load when working on Tier 0 Sovereign (payments, money, orders), Tier 1 Security (auth, tenancy, migrations), or Tier 2 Privacy (PII, erasure) tasks to enforce advanced model tier, Invariant-Breaker adversarial tests, and zero-PII telemetry validation.

中文简介正在准备中。当前显示原文说明。

Load when asked to review a local diff/change set for correctness, regressions, blast radius, affected flows, and missing tests using the knowledge graph; use `review-pr` for intent/checklist/merge-gate evidence.

中文简介正在准备中。当前显示原文说明。

Load for Senior CTO critique, architectural audit, risk review, sequencing correction, or refinement of an existing implementation-plan triad; edits the requested plan/context/tasks in place without creating a worktree or review file. Not for open-ended CTO advice, product implementation, or implementation workspace setup.

中文简介正在准备中。当前显示原文说明。

Load when the user asks to be grilled or stress-tested on a plan, OR when starting Tier 0 (Sovereign), Tier 1 (Security), or Tier 2 (Privacy) tasks with underspecified requirements.

中文简介正在准备中。当前显示原文说明。

Run an offline ASO audit on canonical App Store metadata under `./metadata` and surface keyword gaps using Astro MCP. Use after pulling metadata with `asc metadata pull`.

中文简介正在准备中。当前显示原文说明。

Orchestrate TestFlight distribution, groups, testers, and What to Test notes using asc. Use when rolling out betas.

中文简介正在准备中。当前显示原文说明。

Implement one scoped slice using TDD red-green-refactor. Use when ready to write code for a defined slice of work, or when the user wants test-driven development.

中文简介正在准备中。当前显示原文说明。

Scientific debugging for bugs, flakes, failures, and performance regressions. Use when something is broken, throwing, failing, slow, nondeterministic, or when the user says diagnose/debug this. Builds a trusted repro loop, tests falsifiable hypotheses, installs a regression oracle, and routes durable findings back into ln-* planning.

中文简介正在准备中。当前显示原文说明。

Audit code quality focusing on deep modules, naming, model hygiene, and architectural clarity. Use after a burst of development, when codebase structure needs assessment, or to make code more agent-navigable.

中文简介正在准备中。当前显示原文说明。

Interview the user relentlessly about a plan or design until reaching shared understanding. Use when fleshing out an idea, stress-testing a design, or when the user says "grill me".

中文简介正在准备中。当前显示原文说明。

Author a completed-spec fixture — a seeded intent graph plus a believable four-phase interview transcript — for a brunch feature, so it can feed any downstream tooling that consumes a completed spec or dogfood the UI. Use when you need a real completed specification in .brunch/brunch.db for testing or demo: pick a feature from memory/SPEC.md (or a free description), choose a stress profile (coverage|clean), and produce a registered scenario seed + an interview-questions doc that match exactly.

中文简介正在准备中。当前显示原文说明。

Roleplay a hostile user to find and triage UX pain points.

中文简介正在准备中。当前显示原文说明。

Use for 100-point QA validation of all skills + auto-fix + report.

中文简介正在准备中。当前显示原文说明。

Use when modifying OpenAmer's own core code safely. Gate every change behind the test suite with automatic rollback.

中文简介正在准备中。当前显示原文说明。

Exploratory QA of web apps: find bugs, evidence, reports.

中文简介正在准备中。当前显示原文说明。

Use for A/B experiments on OpenAmer configs and skills.

中文简介正在准备中。当前显示原文说明。

TDD: enforce RED-GREEN-REFACTOR, tests before code.

中文简介正在准备中。当前显示原文说明。

Port-scan, deps, permissions, network, password hygiene.

中文简介正在准备中。当前显示原文说明。

Train an OpenAmer-specific small model from the collected brain data (QLoRA on NVIDIA, CPU smoke-test on laptop). 'The organism becomes its own brain' — build a fine-tune dataset, train, eval, export.

中文简介正在准备中。当前显示原文说明。

Prompt Injection Challenge: LLMail-Inject The competition features 40 unique levels, each combining

中文简介正在准备中。当前显示原文说明。

Symbolic Guardrails for Domain-Specific Agents: Stronger … — 16. Apr. 2026 · We study symbolic guard GAP: OpenHands has a modular SDK design — our too

中文简介正在准备中。当前显示原文说明。

Attack Success Rate (360 runs 60 payloads 2 models ~3 reps) Paradigm gemma4-e2b (local, Ollama

中文简介正在准备中。当前显示原文说明。

LLM) can be manipulated through prompt injection, whether a retrieval pipeline leaks confidential da

中文简介正在准备中。当前显示原文说明。