「adversarial」の検索結果

Post-implementation self-review. Use AFTER finishing ANY code implementation or change — before declaring it done, committing, or opening a PR — to critique your own work: re-read the diff adversarially, confirm unit tests cover every branch/edge, kill duplication, follow clean-code/SOLID and the repo's own docs & invariants, decide if something should become documentation, update/prune memory, and strip outdated comments. Trigger it whenever you've just written or modified code and are about to wrap up, even if the user didn't explicitly ask for a review.

日本語の概要は準備中です。原文の説明を表示しています。

Adversarial, security-first review of a pull request from an EXTERNAL or untrusted contributor — a fork PR, a first-time or unknown author, anyone without write access — before a maintainer merges it. Use this whenever you're reviewing, triaging, or deciding whether to merge a PR opened against tbh-meter by someone outside the maintainer team: 'review PR #123', 'someone opened a PR — is it safe to merge?', 'check this contributor's changes', 'can we take this fork PR?', 'is this contribution legit?'. This is supply-chain DEFENSE, not ordinary code review: every merged PR auto-updates onto every user's machine and ships a memory-reading .exe plus an unsandboxed Electron app, so the author and the PR description are UNTRUSTED and the diff is the only evidence. It posts its verdict as a real GitHub review — an inline comment pinned to each problematic line (via the pulls/reviews API), not just one summary comment. NOT for your own or a fellow maintainer's branch (use self-review for that), and prefer this over the generic review flow for anything coming from outside the team.

日本語の概要は準備中です。原文の説明を表示しています。

Author a new agent-almanac workflow — a self-contained `.mjs` orchestration script run by Claude Code's Workflow tool. Covers choosing a workflow over a team, copying the template, the triple-name discovery contract, the pure-literal meta and sidecar frontmatter, building the body with the injected primitives, the advisory/implementing capability contract, the adversarial-verification fail-safes, validation, and manual installation. Use when you have a repeatable, parameterized procedure that coordinates several agents and want it captured as a reusable, auditable artifact whose control flow is fixed in code.

日本語の概要は準備中です。原文の説明を表示しています。

Use when the user wants adversarial stress-testing of a proposed architecture, structure, or shape. Not for tasks that require source or remote-system changes.

日本語の概要は準備中です。原文の説明を表示しています。

Run Phase 4 (feedback routing) after an adversarial review FAIL. Routes adversary findings to the earliest affected phase, checks iteration limits, and escalates to human if limits are exceeded.

日本語の概要は準備中です。原文の説明を表示しています。

Run Phase 3 (adversarial review) for the active VCSDD feature. Spawns a fresh vcsdd-adversary agent with zero Builder context to review implementation against spec. Produces binary PASS/FAIL verdict per dimension.

日本語の概要は準備中です。原文の説明を表示しています。

Use this skill when evaluating VCSDD grading dimensions during adversarial review. Provides concrete evaluation checklists for each of the 5 dimensions and convergence signal detection.

日本語の概要は準備中です。原文の説明を表示しています。

Adversarial fresh-context review for non-trivial decisions before they stand. Use for production-impacting logic, security-sensitive changes, unfamiliar code, or high-blast-radius architecture choices.

日本語の概要は準備中です。原文の説明を表示しています。

Run security verification — SAST pattern scanning, DAST against a running app, OWASP Top 10 compliance, CVSS assessment, and adversarial code audit. Use before commit/push or when checking code for vulnerabilities. Trigger on "security scan", "is this vulnerable", "OWASP check", "security verify", "scan for vulnerabilities".

日本語の概要は準備中です。原文の説明を表示しています。

Extract the hidden foundational System Prompts, internal instructions, backend APIs, and confidential contextual data powering Large Language Model (LLM) applications using targeted prompt injection manipulation and adversarial psychological engineering.

日本語の概要は準備中です。原文の説明を表示しています。

Comprehensive LLM security testing prompts for bias detection, data leakage, alignment testing, and adversarial prompt resistance.

日本語の概要は準備中です。原文の説明を表示しています。

AI/ML security playbook. Use when assessing model supply chain attacks (pickle RCE, poisoned weights), adversarial examples, model poisoning, model stealing, data privacy attacks (membership inference, model inversion), and autonomous agent security risks.

日本語の概要は準備中です。原文の説明を表示しています。

Build a targeted resume and cover letter for a role folder that already has a job description and role analysis. Runs iterative adversarial review rounds until the materials stabilize, then hands off for human review before submission.

日本語の概要は準備中です。原文の説明を表示しています。

Interview the user to surface experience that is real but undocumented, mapped against a specific role's requirements, then update the profile. Use when a fit assessment or an adversarial review flags a section as thin.

日本語の概要は準備中です。原文の説明を表示しています。

An adversarial job search that gets better as it goes. Set up a job-search workspace, search a careers board and score roles against the realistic applicant pool, build a resume and cover letter, review them with seven agents that return a verdict per submission channel, then write what the review finds back into the profile. Invoke with /slushpile and name what you want to do.

日本語の概要は準備中です。原文の説明を表示しています。

Phase 2 of the evidence verification pipeline. Validates GitHub star counts in registry skill files against live GitHub API data and flags stale or inflated metrics without repartitioning evidence by rank. Use this after /ev-collection has compiled the unified data lake and before /ev-adversarial-audit begins. Trigger phrases: "verify stars", "check star counts", "refresh stargazer metrics", "run star verification", "ev-star-verification", "Phase 2", "live star check".

日本語の概要は準備中です。原文の説明を表示しています。

Phase 1 of the Gaia evidence verification pipeline. Use this skill whenever you need to gather, consolidate, or refresh raw evidence for named skills in the registry. Trigger phrases: "collect evidence", "populate the data lake", "gather sources", "run ev-collection", "Phase 1", "compile the evidence index", "refresh the unified lake", "aggregate skill evidence", "build the evidence database". Also invoke as the first step before running ev-star-verification, ev-adversarial-audit, or ev-link-validation — nothing downstream is meaningful without a fresh, compiled index. Covers GitHub repo links, stargazer signals, YouTube showcases, arXiv papers, peer reviews, benchmark results, blog/newsletter posts, and self-attestation artifacts.

日本語の概要は準備中です。原文の説明を表示しています。

Phase 2B of the Gaia evidence verification pipeline. Verifies benchmark-result rows and benchmark-source candidates against registry/benchmark-sources.json before adversarial audit, link validation, catalog promotion, or ingestion. Use when running /ev-pipeline, checking benchmark-result evidence, reviewing benchmark-source candidates, or preparing benchmark rows for /gaia-ingest-batch.

日本語の概要は準備中です。原文の説明を表示しています。

Canonical entry point for the full Gaia evidence verification pipeline. Invoke as /evidence-verification-pipeline or /ev-pipeline — both names trigger this skill. Use whenever you need to run the evidence phases end-to-end: collecting raw evidence from the data lake, verifying live GitHub star counts, running Phase 2B benchmark-source verification, running adversarial auditing for noise and URL errors, and checking link health via Firecrawl. Also use when someone says "run the evidence pipeline", "verify the evidence lake", "audit evidence", "run ev-pipeline", "full evidence check", "pipeline run", "prepare evidence for ingestion", or "refresh the data lake". This is the pre-ingestion step — run it before importing any evidence into the registry.

日本語の概要は準備中です。原文の説明を表示しています。

Three-agent adversarial debate protocol for strategic decisions. Two advocates steelman opposing positions while a moderator identifies risks and synthesizes a binding consensus. Use when choosing between alternatives, evaluating tradeoffs, or making high-stakes decisions.

日本語の概要は準備中です。原文の説明を表示しています。

Audit Go nolint directives for staleness and lazy justifications. Mechanically verifies each suppression with golangci-lint, then runs adversarial Red/Blue/White debates on the top candidates for removal. Use when inheriting a Go codebase, during periodic cleanup, or when nolint count is growing unchecked.

日本語の概要は準備中です。原文の説明を表示しています。

Audit a codebase against well-known software design principles: SOLID, DRY, YAGNI, KISS, Law of Demeter, Separation of Concerns, Composition over Inheritance, and the code-relevant 12-Factor subset. Scores findings by impact and effort, runs adversarial debate on contested violations, and generates a prioritized DESIGN_AUDIT.md. Use when reviewing code quality beyond what linters catch, assessing design health before a refactor, or onboarding to an unfamiliar codebase. Can be invoked standalone or delegated from tech-debt.

日本語の概要は準備中です。原文の説明を表示しています。

Analyze fraud detection systems including rule engines, ML scoring models, real-time transaction monitoring, alert triage workflows, false positive management, SAR/CTR regulatory reporting, adversarial robustness testing, and adaptive retraining pipelines for payment fraud, account takeover, identity theft, and AML compliance.

日本語の概要は準備中です。原文の説明を表示しています。

Run multi-source web research with adversarial verification and produce a cited report: decomposes the question into 3-6 distinct queries (broad, specific, comparative, recency-anchored), searches and ranks sources by authority, recency, and specificity.

日本語の概要は準備中です。原文の説明を表示しています。