The library

segment-anything-model · Use when you need to segment any object in images using points, boxes, or masks as prompts, or automatically generate all object masks in an image.

aminet-emulator · Use when configuring emulation, managing ROMs, or launching Amiga software.

stable-diffusion-image-generation · Use when generating images from text prompts, performing image-to-image translation, inpainting, or building custom diffusion pipelines.

toss-push · Use for toss-push tasks and operations.

aminet-index · Managing complete Aminet package database lifecycle with incremental RECENT-based updates.

whisper · Robust multilingual ASR on noisy audio; 72.9k GitHub stars, MIT licensed, 680k training hours.

aminet-installer · Safe Amiga package installation with atomic tracking, dependency detection, and enforced scan gates.

knowledge-distillation · Retaining 90%+ of large-model performance in smaller deployable student (70B→7B).

aminet-mirror · Bulk downloading packages with fallback mirrors, atomic state persistence, and resume-from-interruption.

long-context · Processing long documents (32k-128k+ tokens) by extending pre-trained models with RoPE/YaRN interpolation.

aminet-scanner · Enforcing security gates on Amiga package installation via 3-layer scanning (signature + heuristic + emulated).

model-merging · Creating specialized models by blending domain expertise (math+coding+chat) in minutes without retraining.

kp-applied-domain · Guiding generation of Applied & Practical tier content across 10 subjects with real-world relevance.

model-pruning · Achieving 50% sparsity with minimal accuracy loss via one-shot pruning without retraining.

kp-content-authoring · Generating consistent, scaffolded educational content across 4 tiers with prerequisite chains.

moe-training · Scaling model capacity without proportional compute via conditional expert routing.

kp-content-validation · Ensuring all tier packs meet quality bars before publishing.

speculative-decoding · 2-8x inference speedup with small draft model validated by large verifier.

kp-core-academic-domain · Structuring foundational knowledge across Math, Reading, Science, Social Studies, Arts, PE, Tech.

academic-plotting · Creating journal-ready plots with minimal code (IEEE/Nature style templates included).

kp-pipeline-orchestration · Managing multi-tier content workflows ensuring prerequisites before generating downstream tiers.

ml-paper-writing · Rapidly drafting ML papers with automatic figure integration and citation formatting.

kp-quality-review · Filtering low-quality tier packs before release; blocks unmet prerequisites.

presenting-conference-talks · Converting research papers to engaging talks with audience-appropriate complexity scaffolding.

kp-specialized-domain · Route specialized domain queries (philosophy, theology, arts, trades, astronomy) to correct curriculum pack.

systems-paper-writing · Structure a 10-12 page OSDI/SOSP/ASPLOS/NSDI/EuroSys paper with authoritatively grounded paragraph blueprints.

electronics · Route electronics learning queries from Ohm's law through PCB design to correct instructional module.

brainstorming-research-ideas · Move from vague curiosity to concrete, defensible research direction via problem-first vs. solution-first analysis and abstraction-ladder exploration.

google-calendar · Query and modify Google Calendar events programmatically with automatic pagination and empty-field cleanup.

adversarial-pr-review · Verify PR claims match actual implementation against approved spec before merge, detecting scope creep and divergences.

creative-thinking-for-research · Escape local-optimum thinking via bisociation, problem reformulation, and tension-resolution frameworks grounded in decades of creativity research.

commit-changes · Generate conventional, focused commits with context-aware message from current tree state without secrets leakage.

ara-compiler · Ingest papers, code, experiments into falsifiable, agent-traversable knowledge package with cognitive + physical layers and provenance tracking.

create-skill · Author new Hermes skills with correct metadata format and immediate ~/.hermes/ installation for auto-load.

batch-rewrite-pattern · Apply identical refactor (rename symbol, swap import, migrate signature) across many files in single planned batch rather than N manual passes.

ara-research-manager · Manage multi-version research artifact evolution with checkpoints, exploration branching, and job resumption tracking.

octorus · Route incoming queries to most relevant agent skill using trigger pattern matching and confidence scoring.

cartridge-forge · Generate executable parameter cartridges with reproducible hyperparameters and runtime-compatible loader stubs.

ara-rigor-reviewer · Verify ARA claims are grounded in actual source evidence, detect hallucinations, and flag unsourced assertions before publication.

checkpoint-resume-long-job · Enable resumption of long-running jobs with periodic checkpoints rather than full restart on failure.

prompt-master · Create reproducible, version-controlled prompt templates with automated test coverage and performance tracking.

code-review · Perform effort-calibrated code review with bug hunting, reuse/simplification, and efficiency improvements.

commit-style · Retroactively enforce commit style (conventional commits, scoping, subject length) across existing commits.

INSTALL-SKILL · Enable immediate auto-load of skill by installing to ~/.hermes/skills/ without manual activation.

decision-framework-invoker · High-stakes decisions where reversibility matters and second-order consequences are hard to see

.deepschema.SKILL.md.yml · Enforcing behavioral correctness in AI workflow skills without manual gate-keeping

done-retirement · Making done irreversible via push (no local state preservation, no U-turn after Stage 3)

zh-encoding-fix · Recovering garbled Chinese text after write/edit without losing original in backup

execution-grounded-selection · Code selection when semantic voting's 19-52pp improvement over output voting matters

gsd-guide · Reducing hesitation in GSD adoption by previewing what commands will do before running them