“adversarial”的搜索结果

End-to-end Coval adversarial / red-team testing workflow. Builds one adversarial test set (12 core attack vectors plus legitimate controls, each with an expected-behavior checklist), creates a persistent "Adversarial User" persona and a Composite Evaluation metric that scores each scenario against its own expected behaviors, launches a multi-iteration run against the agent (voice or chat), polls for completion, builds a per-scenario pass/fail scorecard, and creates a saved report grouped by Test Case. Use when a user wants to follow the Adversarial & Red-Team Testing cookbook (https://docs.coval.dev/guides/adversarial-red-team-testing) without doing each step by hand. Triggers: "adversarial test set", "red team my agent", "jailbreak / prompt-injection testing", "test my agent against bad actors".

中文简介正在准备中。当前显示原文说明。

Review code changes from three critical perspectives: production failures, maintainability, and security. Get a consolidated report that ranks findings by severity and provides merge guidance, helping examine assumptions that an ordinary self-review may overlook.

中文简介正在准备中。当前显示英语简介。

• Reviewing a pull request before merging
• Getting a second opinion on AI-written code
• Checking authentication or data access code

[omh] Technical proposal facing adversarial scrutiny: independent perspectives attack a proposal, then distill into a bundle a separate planner consumes. Use when the user says: adversarial-consensus, adversarial planning, adversarial plan review, red team this plan, red-team this plan, red team the proposal, multi-perspective review, multiple perspectives.

中文简介正在准备中。当前显示原文说明。

Use this skill when reasoning about generative AI, adversarial machine learning, neural network security, algorithmic fairness, or deep learning fundamentals. This skill channels the thinking of Ian Goodfellow, inventor of Generative Adversarial Networks (GANs). Trigger this skill when the user asks about model robustness, mitigating bias, evaluating AI guardrails, designing generative models, or defending against adversarial attacks. Apply his frameworks of minimax games, adversarial feature learning, and worst-case robustness analysis to shift the user's perspective from average-case optimization to adversarial resilience.

中文简介正在准备中。当前显示原文说明。

Use this skill when reasoning about generative AI, adversarial machine learning, neural network security, algorithmic fairness, or deep learning fundamentals. This skill channels the thinking of Ian Goodfellow, inventor of Generative Adversarial Networks (GANs). Trigger this skill when the user asks about model robustness, mitigating bias, evaluating AI guardrails, designing generative models, or defending against adversarial attacks. Apply his frameworks of minimax games, adversarial feature learning, and worst-case robustness analysis to shift the user's perspective from average-case optimization to adversarial resilience.

中文简介正在准备中。当前显示原文说明。

N-round adversarial review pipeline for empirical research output — the chain from data to LaTeX tables to a manuscript that cites them. A Claude drafter proposes minimal diffs, a deterministic mechanical battery gates every diff from a clean state with a regression gate, a Codex reviewer files check-backed critiques, and a blind judge panel decides residual disputes. Manual-invoke ONLY: trigger when the user explicitly runs /adversarial-empirical-review or names 'adversarial-empirical-review' / 'adversarial empirical review'. Do NOT auto-trigger on generic 'review my results', 'check my tables', or manuscript-editing requests. For prose-style refinement use style-emulation instead; this skill AUDITS WHETHER THE TABLES ARE CORRECT — that each number in the tables is what the analysis code computes, reproduces from the data, and is internally consistent. It is an empirical + code review: the manuscript is read only to resolve table numbering, and prose is not examined.

中文简介正在准备中。当前显示原文说明。

Check training data and machine learning models for malicious changes, including backdoors that cause unexpected behavior on specific inputs. Combine data quality checks, model analysis, and file integrity verification, then document findings and retest after remediation.

中文简介正在准备中。当前显示英语简介。

• Screening third-party data before training
• Checking downloaded model weights before deployment
• Investigating unexpected behavior on specific inputs

Run this skill for Phase 3 of the evidence verification pipeline — the adversarial audit step. Use when you need to check the evidence data lake for bad data before ingestion: dead links, wrong URL formats (tree/ vs blob/), subjective wording ("elite", "high-quality"), stale migration notes, benchmark catalog misuse/vendor-claim leakage, or skills whose star evidence conflicts with classified evidence level. Triggers on phrases like: "audit the data lake", "adversarial check", "ev-adversarial-audit", "check for noise in evidence", "flag bad evidence", "run the audit phase", "quality check the by-type files", or any reference to Phase 3 of the pipeline.

中文简介正在准备中。当前显示原文说明。

Run an adversarial two-phase code review of a change with TWO independent reviewers — Claude and Codex — who review alone, then cross-examine each other's findings, then Claude synthesizes a single weighted verdict. Use when the user says "/adversarial-review", "adversarial review", "review this with Codex", "get Codex to review", "two-reviewer review", "cross-examine this PR", or wants a second independent model to grade a change before merge.

中文简介正在准备中。当前显示原文说明。

本番と同じ失敗を本番前に検出するためのテスト階層設計スキル。 DB方言差、実依存関係、永続化確認、timezone差、packaging整合性、 adversarial regressionをunit/integration/e2e/smoke/packagingに 適切に割り振り、PR CIで最低限の本番同等性を保証する。 Use when designing production-parity tests, closing test blind spots, building smoke suites for real dependencies, verifying persistence beyond UI success, or creating adversarial regression backlogs.

中文简介正在准备中。当前显示原文说明。

This skill should be used when the user wants to "stress-test a design", "challenge an idea", "red team a proposal", "get critical feedback on a design", "validate a design", "adversarial architecture review", "risk assessment", "設計を検証して", "反論をもらいたい", "設計を批判的にレビュー", "デビルズアドボケート", "ストレステスト", "弱点を指摘して", "穴を見つけて", "この設計で良いか", "リスク評価", "設計のアーキテクチャレビュー", or mentions structured adversarial review of a proposal or design. NOTE: Use this for validating PROPOSALS and DESIGNS through adversarial debate, NOT for generic code review, PR review, or normal review requests (use /codex:review or /codex:adversarial-review for those). Also NOT for investigating unknown bugs (use /agent-dialectics:strong-inference for that).

中文简介正在准备中。当前显示原文说明。

Produces an adversarial attack on a legal argument that survives reply. Runs the opposing-counsel discipline twice: an unrestrained first pass, then the same instrument turned on that pass to cut every point that collapses under challenge - attacks on conceded facts, wrong-forum objections, speculation about documents and motives, gotchas with innocent explanations, overclaims, self-refuting assertions, and padding. The deliverable is one standalone attack containing only what can be defended, capped at five heads, not a discussion of the discarded draft. Use to attack, stress-test, red-team or rebut a submission, brief, motion, witness statement, letter or structured legal reasoning. Triggers on "attack this but only with points that hold", "no cheap shots", "what survives reply", "which points can I actually defend", "give me the version I can file", "double-pass adversarial review". Also use when an earlier adversarial review came back overlong or scattershot. Formal British English.

中文简介正在准备中。当前显示原文说明。

Iteratively refine a product spec by debating with multiple LLMs (GPT, Gemini, Grok, etc.) until all models agree. Use when user wants to write or refine a specification document using adversarial development.

中文简介正在准备中。当前显示原文说明。

Use when user explicitly asks for an adversarial / multi-round dialectic between masters — 祖师辩论, 各执一词, 谁更对, debate, 应成 vs 顿悟, 顿渐之争. Differs from /compare-masters (parallel single-round) by being adversarial multi-round via fresh-subagent orchestration. Topics 空有 / 禅净 / 性相 / 戒律 vs 内观 — trigger is adversarial framing: "禅净比较" → compare; "禅净辩论 / 谁更究竟" → here.

中文简介正在准备中。当前显示原文说明。

Use Adversarial Robustness Toolbox (ART) for estimator wrappers, adversarial attacks, defences, poisoning/privacy/extraction, metrics, and certification workflows.

中文简介正在准备中。当前显示原文说明。

Adversarial code review using the opposite model. Spawns 1–3 reviewers on the opposing model (Claude spawns Codex, Codex spawns Claude) to challenge work from distinct critical lenses. Triggers: "adversarial review".

中文简介正在准备中。当前显示原文说明。

Adversarial code review using the opposite model's CLI. Spawns 1–3 reviewers on the opposing model (Codex sessions typically spawn Claude; Claude sessions spawn Codex) to challenge work from distinct critical lenses. Triggers: "adversarial review".

中文简介正在准备中。当前显示原文说明。

Run the weak-agent adversarial test harness against docx-cli. Spawns weak exercise agents (Haiku by default, Sonnet to probe, or a local agent harness's pre-produced runs) to perform real document tasks over six scenarios — five editing (MNDA form-fill + font fidelity, invoice table-edit/restructure + logo replace, résumé styling, contract redlining + commenting, contract finalize via accept/reject + comment reply/resolve) and one authoring (T. S. Eliot poetry journal: multi-column, verse, footnotes, links, figure) — renders every result with Word, has opus judge them against ground-truth rubrics, measures each exercise's tool economy, token cost, wall-clock, and correctness (from transcripts for Claude, the exercise.json ledger for the local harness), and synthesizes a prioritized ergonomics report. Use when the user says 'adversarial review', 'test docx-cli with weak agents', 'run the haiku harness', 'weak agent test', or wants to re-run yesterday's adversarial process.

中文简介正在准备中。当前显示原文说明。

Synthesize the single strongest EVIDENCE-BOUND reviewer case to reject a paper, built ONLY from the evidence ledger (claims.json) + the other auditors' confirmed findings — never free-floating LLM critique. Two fresh cross-model codex threads: an attack writes the ~200-word rejection paragraph (every accusation tagged to an existing claim_id/finding_id), a defense decomposes it and rules each point against the anchored evidence. MEMO-ONLY: emits adversarial-case-builder.memo.md (fed to the adjudicator via --memo) and carries NO verdict weight — tools/adjudicate_findings.py lists it in ZERO_WEIGHT_SKILLS and caps it at info. Honest-null allowed (the paper may survive). Run LAST. Detect-only. Adapted from ARIS kill-argument. Triggers: "adversarial case", "strongest objection", "rejection memo", "kill argument", "最强拒稿点".

中文简介正在准备中。当前显示原文说明。

Multi-pass structural code review with adversarial analysis, scope-aware dispatch, adversarial persona passes, and forced-finding protocol. Supports input modes: branch diff (default), PR number (#123), commit hash, pending changes (--pending). Use when asked to "review this PR", "code review", "pre-landing review", "check my diff", or "review #123". Proactively suggest when the user is about to merge or land code changes. NOT for behavioral verification against a running build (see mk:evaluate); NOT for post-implementation simplification (see mk:simplify).

中文简介正在准备中。当前显示原文说明。

Full adversarial quality loop — implement, self-attack, parallel verification, quality gates, final validation. 全品質循環：實現、自攻、並行驗證、質量關卡、最終確認. Use when: implement task with quality, run adversarial review, verify code quality, run quality gates, complete workflow task

中文简介正在准备中。当前显示原文说明。

Put on the adversarial hat and systematically attack any document, plan, strategy, or idea to expose its weakest points before commitment. Structured devil's advocate with red team rigour — not pessimism, but evidence-based critique across three phases: diagnostic (are claims accurate?), creative (is the problem artificially constrained?), challenge (are solutions robust?). Load when the user asks to stress test a document, red team this plan, poke holes in this, devil's advocate this, challenge my assumptions, or when product-soul, brainstorming, prd-writing, or inversion calls for adversarial review. Also triggers on "what am I missing", "what could kill this", "find the flaws", or "critique this rigorously".

中文简介正在准备中。当前显示原文说明。

Analyze a Coval adversarial / red-team testing report and turn it into an agent-hardening plan. Use when a user provides a Coval report URL, report export, run IDs, screenshots, or a per-scenario scorecard from an adversarial sweep and wants evidence-backed next steps such as prompt/guardrail changes, refusal hardening, verification fixes, escalation routing, or expanded attack coverage.

中文简介正在准备中。当前显示原文说明。

Uses multiple agents with different critical roles to challenge assumptions, complexity, risks, and alternatives. Findings that withstand review go to a dedicated planner, who turns them into sequenced tasks with explicit success criteria.

中文简介正在准备中。当前显示英语简介。

• Reviewing assumptions before implementation
• Checking integration risks and edge cases
• Comparing alternative approaches