AI & ML

sboghossian-mini-claude-for-legal Skill

Use when migrating a French cookie policy (politique de cookies) drafting or review skill into the mini-claude-for-legal format. The adapter maps French CNIL cookie-compliance logic — consent requirements for non-essential cookies, exemption categories, cookie-banner specifications, and retention limits — into the standard skill model. Primary jurisdiction France; relevant for any EU/EEA entity targeting French users.

sboghossian-mini-claude-for-legal Skill

Use when migrating a skill-creation or skill-authoring tool originally built for the OpenAI API into the mini-claude-for-legal format. The adapter maps legacy OpenAI skill-generation prompts — GPT-4 function-calling schemas, structured output formats, and skill-templating pipelines — into the standard SKILL.md format for the open-source repository. Triggers when importing any OpenAI-native prompt-engineering or skill-factory workflow.

yongjianwan Bundle

AI Agent Skill for meeting management. Audio transcription and structured storage with entity linking.

drixxodev Skill

Reference knowledge for Notion workspace triage — workspace structure, client routing rules, TOC format, folder conventions, and MCP tool mapping. Use when triaging root pages, creating client folders, or managing the workspace hierarchy.

autohandai-community-skills Bundle

Use to help users get started with Nemo Gym reward profiling. Covers the basic ng_run, ng_collect_rollouts, and ng_reward_profile workflow, repeated rollouts, materialized inputs, rollout JSONL artifacts, task and rollout identity, output inspection, partial profiling, and rollout_infos. For failed jobs, prefer nemo-gym-debugging.

autohandai-community-skills Bundle

Use when adding, modifying, optimizing, or debugging CuTile autotuning code. Trigger signals: `exhaustive_search` / `replace_hints` / `hints_fn` / `cuda.tile.tune` in code, `autotune` in filenames, or correctness/performance issues in autotuned CuTile kernels. Covers: tune-once/cache/launch pattern, per-architecture configs (sm80–sm120), parameter space design (tile sizes, occupancy, num_ctas), and 7 common pitfalls with solutions.

sboghossian-mini-claude-for-legal Skill

Use when validating the consistency of the subscription data model in a legal AI SaaS product — specifically the relationships between Stripe customers, users, tenants, subscriptions, plans, feature flags, invoices, and usage records. Flags orphaned records, duplicate active subscriptions, plan/feature-flag mismatches, and proration inconsistencies, and routes findings to the appropriate billing or database owner.

autohandai-community-skills Skill

Orchestrate multi-agent legacy modernization assessment, planning, and code-conversion workflows.

autohandai-community-skills Bundle

Evaluates accuracy of quantized or unquantized LLMs using NeMo Evaluator Launcher (NEL). Triggers on "evaluate model", "benchmark accuracy", "run MMLU", "evaluate quantized model", "accuracy drop", "run nel". Handles deployment, config generation, and evaluation execution. Not for quantizing models (use ptq) or deploying/serving models (use deployment).

gptomics Bundle

Discovers, classifies, and masks repetitive elements and transposable elements with RepeatModeler2 (de novo family library), RepeatMasker (masking against a library), EDTA (plant/structural TEs), or EarlGrey (auto-curating wrapper), and quantifies TE expression from RNA-seq with TEtranscripts/SQuIRE. Covers de-novo-library-as-curation-project, soft-vs-hard masking, the domesticated-gene over-masking massacre, Dfam-vs-RepBase, TE classification (Class I/II, family-vs-copy), Kimura repeat landscapes, LAI, and the RNA-seq multimapping problem. Use when masking repeats before gene prediction, building a TE library for a non-model genome, or analyzing transposable-element content or expression.

gptomics Bundle

Build model-corrected evolutionary distance matrices and distance trees (NJ, BIONJ, FastME, UPGMA) with Biopython Bio.Phylo plus R ape/phangorn/FastME. Covers why a distance is a model-corrected estimate of substitutions per site that undercounts raw because of multiple/back/parallel hits (saturation); why the matrix discards the per-site information ML keeps; the LogDet/paralinear fix for compositional heterogeneity; the UPGMA molecular-clock trap; and the Bio.Phylo landmine that DistanceCalculator offers only identity/matrix distances, not JC/K80/TN93. Use when computing a distance matrix, building a fast NJ/FastME tree, seeding an ML search, barcoding, or testing substitution saturation before a deep tree. Routes ML and starting-tree work to modern-tree-inference, alignment quality to alignment/alignment-io, and tree I/O to tree-io.

gptomics Bundle

Infers maximum-likelihood phylogenetic trees with IQ-TREE2 and RAxML-NG -- model selection (ModelFinder), branch support (UFBoot2, SH-aLRT), concordance factors (gCF/sCF), partitioning, topology tests, and long-branch-attraction control. Covers why an ML tree inherits every flaw of the assumed model and the fixed alignment, why reported support measures repeatability under resampling and not correctness, why UFBoot uses a >=95 cutoff and not the bootstrap-70 rule, and why a node with UFBoot 100 but gCF ~35 is essentially unresolved ILS rather than a clade. Use when inferring an ML tree, selecting a substitution or partition model, choosing or interpreting support measures, testing an a-priori topology, or diagnosing LBA. Routes model-free distance trees to distance-calculations, posteriors to bayesian-inference, and species trees under ILS to species-trees.

sboghossian-mini-claude-for-legal Skill

Use when selecting an embedding model for semantic search, document similarity, or retrieval-augmented generation (RAG) in a legal AI product. Covers the criteria specific to legal text (multilingual Arabic/English, domain vocabulary, long-document chunking), a comparison of leading embedding models, and configuration recommendations for legal retrieval pipelines serving MENA and common-law jurisdictions.

sboghossian-mini-claude-for-legal Skill

Use when configuring or invoking the LLM-as-judge evaluation system that scores legal AI outputs against multiple rubrics. Provides the canonical system prompt template, JSON output schema, ensemble configuration, and bias mitigation rules for automated quality scoring of legal AI responses.

sboghossian-mini-claude-for-legal Skill

Use when migrating a GDPR privacy-notice drafting or review skill originally built for the EU context into the mini-claude-for-legal format. The adapter maps legacy notice templates, Article 13/14 disclosure checklists, and layered-notice structures into the standard skill model, with support for French (CNIL), UK (ICO), and cross-border EDPB guidance. Also relevant for UAE PDPL and Lebanon privacy-notice equivalents.

sboghossian-mini-claude-for-legal Skill

Use when migrating a security-review skill originally built for the OpenAI API into the mini-claude-for-legal format. The adapter maps legacy security-analysis logic — threat modelling, vulnerability classification, OWASP alignment, data-protection risk flags, and remediation recommendations — into the standard skill model. Relevant for legal tech products, API security assessments, and AI-system due-diligence workflows.

sboghossian-mini-claude-for-legal Skill

Use when parties evaluating a potential transaction or collaboration need a mutual non-disclosure agreement covering standard confidentiality obligations, permitted disclosures, term, and governing law. Addresses the key MENA-specific traps in NDAs — enforceability under UAE Civil Law, the treatment of written form requirements, language obligations in KSA, and the distinction between bilateral and unilateral NDAs in common-law vs. civil-law practice.

autohandai-community-skills Bundle

Configuration conventions for NeMo-RL. YAML is the single source of truth for defaults. Covers TypedDict usage, exemplar YAML updates, and forbidden default patterns.

autohandai-community-skills Bundle

Check whether AutoDeploy YAML configs were actually applied by analyzing server logs and optionally graph dumps (AD_DUMP_GRAPHS_DIR). Use when the user wants to verify config application, debug config issues, or check if AutoDeploy transforms (piecewise CUDA graph, multi-stream, sharding, fusion, etc.) were applied or fell back. Triggers on: "check config", "verify config", "ad-conf-check", "were my configs applied", "config not working", "check if piecewise is enabled", "check log for config", or any request to compare AD YAML settings against runtime behavior.

autohandai-community-skills Bundle

Enable and interpret TensorRT-LLM AutoDeploy FX graph text dumps via AD_DUMP_GRAPHS_DIR. Use when you need before/after graphs per transform, to locate subgraphs, or to confirm a rewrite ran. Paths and behavior are grounded in tensorrt_llm/_torch/auto_deploy (GraphWriter, BaseTransform). Complements ad-add-fusion-transformation.

autohandai-community-skills Bundle

Performance analysis coordination workflow. Guides profiling delegation, bottleneck classification (compute/memory/launch/communication/sync), and structured report generation. Use when the user asks to analyze performance, profile a workload, check MFU/SOL, or diagnose bottlenecks.

gptomics Bundle

Differential abundance (DA) and differential state (DS) analysis for flow and mass cytometry - tests which cell populations change in frequency or marker expression between conditions using diffcyt (edgeR/voom/GLMM for DA, limma/LMM for DS), with cydar, CITRUS, and compositional methods (sccomp, scCODA, DCATS) as alternatives. Covers the sample-is-the-experimental-unit principle, design/contrast and mixed-model formulas, compositionality of cluster proportions, and FDR across clusters. Use when comparing populations between groups, choosing a DA method, handling paired/batch designs, or deciding whether compositional correction is needed.

gptomics Bundle

Calls germline small variants (SNPs and indels) from Oxford Nanopore and PacBio HiFi long reads with Clair3, a two-stage (pileup + full-alignment) deep-learning caller, selecting the chemistry- and basecaller-version-matched model, enabling read-based phasing, and benchmarking against GIAB with stratification. Covers why the model string is the experiment (no auto-detection, silent degradation on mismatch), why ONT homopolymer/STR indels are the residual error whole-genome F1 hides, and the somatic/trio/RNA boundary to the ClairS/Clair3-Trio family. Use when calling germline SNVs/indels from ONT or HiFi BAMs, choosing a Clair3 model, phasing variants, or benchmarking long-read calls.

sboghossian-mini-claude-for-legal Skill

Use when running or extending the adversarial prompt benchmark dataset that tests the legal AI system's robustness against jailbreaks, out-of-scope requests, unauthorized-practice attempts, privacy violations, and hallucination bait. This dataset catches the most expensive failure modes and must be run on every model deployment.

sboghossian-mini-claude-for-legal Skill

Use when running the legal research benchmark that tests statute lookup, case law retrieval, multi-jurisdictional comparison, and regulator guidance queries across DIFC/ADGM/UK/FR/MENA. Hallucination rate is the primary metric — any fabricated citation is an automatic fail. Contains 30 prompts with a mandatory <1% hallucination target.

sboghossian-mini-claude-for-legal Skill

Use when migrating a Lawvable Outlook email-processing skill into the mini-claude-for-legal format. This adapter maps the legacy Lawvable email-ingestion configuration — Microsoft Graph API authentication, email threading, legal-matter tagging, privilege flagging, and action-item extraction — into the standard skill model. Triggers on import of any Lawvable-native Outlook integration workflow.

sboghossian-mini-claude-for-legal Skill

Use when migrating a Lawvable tabular-review skill into the mini-claude-for-legal format. The adapter maps Lawvable's structured table-based legal review output — clause-by-clause issue tables, risk-column scoring, and side-by-side comparison formats — into the standard skill model. Triggers on import of any Lawvable-native tabular review or comparative analysis workflow.

gptomics Bundle

Detect syntenic blocks and structural rearrangements between genomes using MCScanX (Wang 2012), JCVI/MCScan (Tang 2008 Python), GENESPACE (Lovell 2022) for orthology-anchored riparian visualization, SyRI for structural variation, AnchorWave for sequence-level synteny, i-ADHoRe 3.0 for highly diverged species, SynNet for synteny networks, and ntSynt for multi-genome macrosynteny. Use when identifying collinear gene blocks across species, distinguishing macrosynteny from microsynteny, detecting inversions/translocations/duplications, anchoring orthology in WGD lineages, producing publication riparian plots, computing synteny block age via Ks (cross-references whole-genome-duplication), or running synteny-aware ortholog inference in polyploids.

gptomics Bundle

Corrects the gene-independent copy-number artifact in CRISPR-Cas9 screens (Aguirre 2016 / Munoz 2016 Cancer Discov) where amplified loci appear essential from DNA-damage burden of simultaneous cuts. Covers the gene-independent DNA-damage / G2-arrest mechanism, CRISPRcleanR (Iorio 2018) unsupervised pre-hoc correction, CERES (Meyers 2017) joint CN + gene-effect model, Chronos (Dempster 2021) DepMap-standard population-dynamics + CN model with lowest residual bias, the decision tree by data availability, the Spearman LFC-vs-CN diagnostic, focal-amplification examples (ERBB2 in HER2+, MYC in colorectal, FGFR1 in head and neck), and CRISPRi/a alternatives that bypass the artifact. Use when screening cancer cell lines, diagnosing essentiality at amplified loci, choosing CRISPRcleanR / CERES / Chronos, deciding whether CN correction is needed before MAGeCK / BAGEL2 / drugZ, or switching from Cas9 to CRISPRi.

gptomics Bundle

Analyzes time-series and longitudinal RNA-seq for differential expression and trajectory structure. Covers DESeq2 LRT with reduced models, time as factor vs continuous vs natural splines, maSigPro (Nueda 2014 for RNA-seq), ImpulseDE2 with explicit impulse-model failure modes, DREAM for repeated measures via linear mixed models, pseudoreplication avoidance, conditional vs marginal modeling, and trajectory clustering with DPGP, Mfuzz (with Schwämmle 2010 fuzzifier estimation), and splines+k-means. Use when modeling time-course or longitudinal expression, choosing factor vs spline, handling repeated measures from the same subject, avoiding pseudoreplication, clustering temporal trajectories, or selecting between dedicated time-course tools and pairwise+LRT.

autohandai-community-skills Bundle

Testing web applications for clickjacking vulnerabilities by assessing frame embedding controls and crafting proof-of-concept overlay attacks during authorized security assessments.

sboghossian-mini-claude-for-legal Skill

Use when a lawyer needs to draft a formal demand letter asserting a client's claim against a counterparty — stating the legal basis for the claim, damages suffered, supporting evidence, and a deadline for payment or response before legal proceedings are commenced. Applicable across MENA (UAE, KSA, LB, EG), DIFC/ADGM, and any specified jurisdiction; addresses the different formal requirements for pre-litigation demand in civil-law and common-law MENA contexts.

sboghossian-mini-claude-for-legal Skill

Use when running the multilingual benchmark that tests language detection accuracy, output language matching, Arabic legal terminology quality, and bilingual document formatting across English, Arabic, French, and mixed-language inputs. Key metric is language-match rate ≥ 95%.

gptomics Bundle

Tests whether two genomic interval sets overlap (colocalize) more than expected by chance using a permutation test against a structured-genome null model. Covers bedtools fisher (analytic 2x2 screen), bedtools shuffle + jaccard permutation, GAT (isochore/GC-conditioned simulation with FDR), regioneR (flexible permutation, randomizeRegions vs circularRandomizeRegions, localZScore), LOLA (universe-relative Fisher against a region database), and GREAT/rGREAT (regulatory-domain binomial + hypergeometric for ontology-from-regions). Stresses the universe/background choice, matched background, blacklist exclusion, and multiple-testing control. Use when asking whether peaks/regions are enriched at enhancers/TFBS/features, scoring region-set colocalization or region-set enrichment, comparing CNV/SV concordance, or turning an overlap count into a defensible p-value.

gptomics Bundle

Performs proximity operations on genomic intervals with bedtools (closest, window, flank, slop) and pybedtools - nearest-feature queries with signed/strand-aware distance, fixed-radius window searches, strand-aware promoter construction, and interval extension. Covers the closest -d/-D a/b/ref/-t/-k/-io/-iu/-id flags, the -D ref strand sign-flip, silent chromosome-end clipping in slop/flank, -t all tie double-counting, and the critical distinction between a geometry answer (nearest TSS) and a biology answer (which gene an element regulates). Use when assigning peaks or variants to genes, defining promoters from a gene model, building distance-to-TSS distributions, finding features within a window, or extending intervals - and when deciding whether nearest-gene is a fair prior (GWAS locus) or a trap (distal enhancer).

gptomics Bundle

Polishes Oxford Nanopore draft assemblies to higher consensus accuracy with medaka, a basecaller-model-specific neural consensus net, produces haploid variant calls (VCF) for microbial, mitochondrial, or viral samples, and generates amplicon/viral consensus sequences. Covers the model-matching footgun that silently degrades output, why Racon-first is obsolete and medaka runs directly on Flye output as a single pass, why HiFi must never be fed to medaka, the v1->v2 subcommand renames, and the precise medaka_variant deprecation. Use when polishing an ONT-only assembly, generating an amplicon/viral consensus, calling a haploid ONT consensus, or deciding whether medaka, dorado polish, or Clair3 is the right tool.