Skills by orionshaowswmw

Detect and repair partially wiped agent workspaces with integrity checks, signed manifests, guarded restore recipes, bounded local recovery state, and explicit off-box sync. Use when files, scripts, trees, models, or build outputs disappear or lose integrity between turns.

Seven offline mechanisms against slow/stale/zombie/sycophantic agent turns: prompt compaction, request fencing, zombie detection, CAPTCHA triage, anti-sycophancy spine, delivery register, invention quarry. Use when chat feels laggy, reconnects surface old answers, long chats degrade, or the agent caves under contradiction. JSON contracts; state per-agent under ~/.arena_turn; no network, no sudo.

Quota-aware LLM router that squeezes maximum usable AI out of free-tier API keys across Gemini, Mistral, OpenRouter, Kilo and Cerebras plus any OpenAI-compatible endpoint (including local Ollama/llama.cpp/vLLM). Probes every model on every key, measures real quality and real published rate limits, then routes each request to the cheapest model that can do the job — spending abundant capacity first and reserving scarce daily quota for when it is actually needed. Persists cooldowns to disk so a 429 discovered in one process is respected by the next. Use when an agent must make many LLM calls on free keys without hitting rate limits, when "all models failed", or when deciding which of several provider keys to use for a task.

Opt-in, model-neutral guidance for evidence-aware, dignified AI communication, with a compact response contract and offline deterministic text audit. It never injects prompts, edits host configuration, calls networks, reads secrets, or treats heuristics as truth.

Iran Chemical Database — live, dated, auditable, BEST-EFFORT index of chemical offerings in configured public Iranian supplier catalogues (websites + public Telegram channels). HTTrack/WooCommerce-REST/Telegram mirroring → local-only parsing → RDKit/PubChem/CAS-validated PostgreSQL with FastAPI + Streamlit. Fail-closed Iranian-suppliers-only country gate; coverage measured and published, never claimed complete. Installation = software + queued crawl, not a populated dataset. Ships a 1399-molecule CID-unique confirmed-organic seed baseline (v2.22, 2026-08-27: v2.19 primary + live Telegram/WooCommerce/sitemap crawl + 5-model fleet normalization, every new identity PubChem-confirmed). For academic procurement research.

Model-agnostic, agent-agnostic fidelity-first pipeline converting operator-authorized Persian/English RTL lecture PDFs into offline HTML study guides — recall-first dual OCR (PyMuPDF + Tesseract fas+eng PSM ensemble), rendered-page evidence, multi-model correction, session-grounded enrichment (tables/flashcards/quizzes/mnemonics/summaries/scenarios), measured fidelity, QA gates, verified ZIP. v1.5.0 runs on ANY model family through 8 API dialects (OpenAI, Responses, Gemini, Anthropic, Cohere, Ollama, HuggingFace, offline mock) or with no model at all, auto-discovers providers from the host agent's environment, self-heals provider quirks and model retirements, and exposes one deterministic CLI/MCP entrypoint plus cross-model consensus so different agents reproduce the same intended result.

Anti-stuck/anti-snapshot-wipe guard for agentic sandboxes with actual selfheal_runner.sh library, byte-verified GGUF manifest, native CPU rebuild +7-10%, har...

Defensive multi-agent SRE/SecOps red-team and purple-team resilience commander with working mode selection, command validation, approval gates, and a machine-readable model quality-floor matrix. Use when planning authorized incident response, defensive red-team/purple-team exercises, model-resilience fallbacks, rollback planning, or evidence handling. Defensive-only, authorization-gated, non-offensive; no attack traffic, no login bypass, no credential collection.

Four-agent image-prompt swarm (Muse drafts, Guardian safety-checks, Critic perfects realism, Composer finalizes) that produces lawful, consent-aware, non-deceptive symbolic imagery prompt packs as validated JSON for journalism, protest, censorship and free-expression themes. Use when the operator asks for a symbolic image or image prompt about protest, censorship, propaganda, press freedom, or free expression.

Privacy-first, lawful connectivity troubleshooting for the Iran context (FA/EN) — use when diagnosing internet/service outages, Wi-Fi or mobile-data failures, DNS/TLS/cert problems, ISP last-mile issues, blackout continuity, or preparing redacted evidence, bilingual support tickets and low-bandwidth reports. Ships working scripts (PII redactor with Iranian formats + checksums, defensive guard, 60-second triage, status-page rate limiter, owned-scope diagnostics, ≤100KB report builder) and 55 templates. No bypass, no scanning, no credential collection — ever.

Offline, defensive robustness auditor for LLM benchmarks: n-gram exact + shingle-Jaccard paraphrase contamination, temporal pre/post-cutoff gaps, TS-Guessing above-chance detection, option-letter selection bias (chi2), few-shot curve noise, LLM-judge position/verbosity/rubric-echo bias and hidden-instruction payload detection, paired McNemar + Wilson + deterministic bootstrap for score comparisons, WORKED mitigations (permutation majority ensemble, blind content normalization), documented 0-100 severity formula, hash-chained per-target history ledger with trend deltas. Findings cite a STATIC 17-id exploit catalogue with explicit computable flags — invisible exploit classes are declared, never fabricated. 100% stdlib python3. NO network, NO telemetry. Defense/auditing only.

Find evidence-bounded candidate research gaps and output a ranked, source-linked report with a candidate research question per gap. Uses a stdlib-only CLI over key-free scholarly APIs (OpenAlex, Semantic Scholar, Crossref, Europe PMC, PubMed, arXiv) plus manual human-in-the-loop guidance. Grounded in PICOS, AHRQ/Robinson gap reasoning, a six-type taxonomy, a five-dimension rubric, reproducible search provenance, and an explicit anti-confabulation protocol.

Executable power-user playbook for arena.ai. Use when choosing an arena.ai mode (Direct / Agent / Side-by-Side / Battle) for a task, reading or checking the Agent Arena leaderboard, screening a response with measurable weak-response flags, chunking long Agent work with SESSION-STATE.md carry, or falling back to cloud providers when arena.ai throttles or is down. Bundles a dated, sourced model snapshot (2026-09-05) plus offline python3-stdlib scripts: mode advisor, weak-response screener, leaderboard rotation checker, state manager, local feedback log.

Build defensible molecular ML for drug discovery — QSAR, virtual screening, ADMET, toxicity, binding affinity, drug-target models — with Random Forest, SVM/SVR, and Gradient Boosting, plus dataset analysis, inductive conformal prediction, multi-model consensus, diversity-aware batch selection, and cheminformatics-aware interpretation. Use for curating bioactivity data, comparing RF/SVM/XGBoost, leakage-resistant validation, library screening, calibrated uncertainty, disagreeing-model reconciliation, diverse assay-batch selection, feature-to-molecule mapping, applicability-domain assessment, or auditing a molecular ML paper.

Battle-tested runbook for running long (100 ns) OpenMM molecular dynamics on Kaggle's free GPU (P100 sm_60) — covers the mandatory OpenMM 8.3.1 pin, checkpoi...

Docks small molecules against human pancreatic lipase (PDB 1LPB, lipase+colipase+Ca2+) across 5 validated sites with AutoDock Vina — pH 7.4 protonation, tautomer/stereoisomer enumeration, multi-seed consensus, native re-dock RMSD gate, and calibration drift detection. Runs locally or on free Kaggle CPU kernels. Use when the user asks to dock, screen, or rank compounds against pancreatic lipase, PNLIP, hPL, or for anti-obesity/lipase-inhibitor virtual screening.

Authorization-gated coordination skill for agents that need to plan, verify, and troubleshoot a DNSTT client/server link without sharing private keys, scanning resolvers, changing DNS/firewalls, or auto-executing tunnel commands. Uses short-lived rendezvous cards, out-of-band public-key fingerprints, optional HMAC authentication, loopback-by-default services, deterministic argv plans, and secret-free status handoffs. Use only on domains, resolvers, servers, and services the operator owns or is explicitly authorized to administer.

哈希钉扎的 agent 工作区重建验证：纯标准库、离线、确定性 CLI。分类"良性质漂移 vs 真损坏" （$(cat) 尾部换行剥离 / CRLF / 截断粘贴 / HTML 错误页 / 同尺寸改动），批量清单校验， runbook 钉扎提取，CommonMark 正确的步骤提取（内嵌围栏不再静默截断），快照擦除后 状态判定与步骤路由。JSON 机器可读输出，每条带 next_action。不修改任何被验证文件。

CyberScope — 可搜索的公开网络攻防/监控/审查方法参考目录：10 类目、62 方法、83 条 公开来源（MITRE ATT&CK、CISA、NIST、EFF、OWASP、SANS 等）。零依赖 Python 标准库 离线 CLI：加权搜索（title>keywords>description>resources）、单条明细、导出 json/csv/md、来源静态验证、目录质量报告（自改进钩子）、校验和锚点。纯参考性 文档——不含操作性/利用步骤。只读，不联网，确定性输出。

Turn an AI agent into an expert at inference — difficulty triage, deliberation budgets, self-consistency, claim-level verification, citation gates, judge ensembles, calibrated abstention, cost cascades, semantic caching, deep-research loops, agent memory, MCP exposure, eval regression gates, and answer certificates. Use for deciding how much inference compute to spend, verifying claims before asserting them, avoiding overthinking/hallucination/verification-theater, and controlling inference cost.

Adds structured JSON logging, error classification, retry with exponential backoff and full jitter, circuit breaking, performance profiling, state capture, and auto-healing to any skill or script, in Python and Bash. Use when a skill or agent needs debugging, error recovery, resilience against flaky network or rate-limited calls, thundering-herd-safe retries, crash diagnostics, performance profiling, memory monitoring, or a reproduce-diagnose-fix-verify workflow for an existing bug.

Evidence-first, offline tuning of llama.cpp GGUF inference on CPU and constrained edge hosts. Inspects CPU topology, renders compatible benchmark plans, optionally runs an explicitly supplied local llama-bench, ranks pp/tg/pg results with variance-aware confidence, verifies output gates, and renders (never executes) a measured command. It does not install llama.cpp, download models, call APIs, or claim universal defaults.

Persist a deterministic skill index into the agent system prompt. One stdlib-only CLI (index/prompt/inject/verify/stats/hook), idempotent marker-block injection, no truncation, offline.

Cross-model intellect relay across a mesh of stored API keys (multi-provider, multi-key). The conductor farms work out to a choir of models reachable via EVERY stored credential — each provider's api_key and each alt_api_keys entry is its own health-tracked voice with automatic per-key failover — and models that need help are answered using other providers' keys (e.g. a Mistral-key model that is rate-limited or unsure consults Gemini and llm7.io keys). Core moves: (1) weak/cheap models sing candidates in parallel and a reasoning model aggregates them (Mixture-of-Agents, arXiv:2406.04692); (2) vision relay — text-only callers hand images to vision-capable voices (arXiv:2502.16428); (3) mandatory escalation — low confidence, errors or rate limits route to a reasoning court from other providers (FrugalGPT cascade, arXiv:2305.05176); (4) judge panels score 0-10 and a persistent reward ledger punishes weak voices while key-health cooldowns rotate load (arXiv:2306.05685). Plus self-consisten

Enhanced Non-Blocking Agent Execution with AI-powered improvements. Prevents "agent stopped responding / stuck / no output" failures in sandboxed agent runtimes (Arena Agent Mode, OpenClaw, Codex) where a single long tool call blocks the turn and stdin is closed so interactive prompts hang forever. NEW in v2.0.0: - Full jobctl.sh implementation with all commands - Multi-model compatibility (OpenAI, Anthropic, Mistral, Groq, etc.) - Token usage optimization and monitoring - Hallucination reduction via verification - Self-improving through feedback loops - Comprehensive debugging and logging - Idempotent operations and graceful error handling - Watchdog timers for timeout protection - Callback URL support for async notifications - Durable state persistence across turns Provides the detach → bounded-poll → durable-state pattern plus a ready-to-use jobctl.sh runner with enhanced features for production use.

Offline website mirroring with HTTrack — snapshot one page (with its assets) or mirror a bounded site to disk for offline reading, backup, or research. Provides scripted recipes (doctor/snapshot/mirror) with polite defaults (robots=always, 2 sockets, depth-limited), strict URL validation, resumable mirrors, and stable JSON reports for agents.

Offline SBOM + verifiable compliance scanner for agent projects: generates a CycloneDX 1.5 JSON SBOM from npm/pip/go manifests, audits the tree with a curated rule engine (committed-secret detection with redacted evidence, unpinned-dependency supply-chain check, declared-vs-actual network egress drift, model-artifact provenance, license presence), emits findings mapped to EXPLICIT control refs (OWASP LLM Top 10 2025, OWASP Agentic AI, NIST SSDF SP 800-218, CISA minimum SBOM elements), filters policy by severity, keeps a hash-chained tamper-evident audit trail, and trends findings across runs for self-improving posture. 100% stdlib python3. NO network, NO telemetry.

BOM (bill of materials) vulnerability scanner that parses 8 lockfile formats, matches versions against an OSV-verified advisory database, and emits provenance-stamped findings. Use when auditing dependencies of an authorized project for CVEs, building an SBOM, assessing supply-chain risk, or generating a machine-readable security report. Offline by default; explicit online mode queries api.osv.dev with name+version only.

解析复杂医学 DICOM 文件（.dcm）：纯标准库、离线、确定性。读取 Implicit/Explicit VR LE 元数据与序列，导出未压缩像素为 PNM，一致性检查，PS3.15 基础配置子集去标识化， 生成确定性合成测试文件。压缩像素（JPEG/JPEG2000/RLE/MPEG）诚实检测并指向 pydicom+pylibjpeg，绝不猜测像素值。仅技术检查，不用于诊断。