Total 56,288 skills, AI & Machine Learning has 9373 skills
Showing 12 of 9373 skills
Provides expertise on Chroma Cloud integration for semantic search and hybrid search applications. Use when the user is working with Chroma Cloud, CloudClient, managed collections, Schema(), Search(), hybrid search, or Chroma Cloud CLI workflows.
Stage 4 of the Clinical ASR Flywheel. Use when priority KER is above 0.3 to run stock NeMo SFT on Parakeet TDT v2 and offline cycle N+1 re-eval. NOT for generic word boosting (use /finetune-asr).
Techniques for reducing peak GPU memory in Megatron Bridge — expandable segments, parallelism resizing, activation recompute, CPU offloading constraints, and common OOM fixes.
Agent Platform Model Registry Management. Use when you need to upload, list, describe, update, or delete machine learning models (and their versions) in the Agent Platform Model Registry. Don't use for model training, model deployment to endpoints, or managing non-Agent Platform models.
Self-healing browser automation framework that connects LLM agents directly to Chrome via CDP. Use when the user needs autonomous browser tasks, clean browser verification, Codex or Antigravity browser control, Claude-safe screenshots, adaptive helper code in `agent_helpers.py`, domain skills, or Browser Use Cloud escalation. Triggers on: browser-harness, self-healing browser, llm browser automation, cdp agent, chrome devtools agent, codex browser automation, antigravity browser automation, claude screenshot error, claude image error, agent browser task, browser-use harness, domain skills browser.
Read-only Jetson health snapshot for identity, memory, GPU, thermal, power, storage, services, and top processes.
Portable AI SDLC shared-helper runtime. Use when an AI assistant installs, verifies, diagnoses, or repairs project-scoped AI SDLC skills whose deterministic scripts depend on shared state, artifact, context, path, TOON, migration, or index modules. This is an installation dependency, not a lifecycle entry point.
AI SDLC context-aware navigation workflow. Use when an AI assistant needs to determine what to do next, select the right installed skill, start or resume a feature, explain blockers, inspect available capabilities, or provide evidence-backed required and optional next actions from repository state. Supports `--quick-flow` for compact guidance and `--full-flow` for stricter context verification.
Use the local `5dive` CLI on a 5dive runtime VM to spawn, inspect, send to, and tear down sibling agents. Trigger when the user wants a worker, sub-agent, side task, parallel run, fan-out, or to delegate — or names a sibling agent ("ask X", "ping X", "tell X", "hand off to X", "coordinate with X"); confirm it exists via `5dive agent list --json`, then `agent send`. Also for inspecting/restarting/pairing an existing agent, a machine-readable health check (`5dive doctor --json`, `5dive selfcheck --json`), a task's causal history (`5dive trace <id|DIVE-N>`), the current model id per alias (`5dive models`), the host-shared task queue + org chart (`5dive task`, `5dive org`), grouping a multi-task effort under a project (`5dive project add`, `task add --project`), recurring/scheduled work (`task add --recurring`, `5dive heartbeat`), parking a question on a human (`task need`, risk-tiered via `--tier`) or snoozing work (`task park --wake`), searching the team's accumulated memory/wiki (`5dive memory search`) or compiling a durable one into it (`5dive memory add`), reading fleet health / token burn / the daily standup (`5dive supervisor`, `5dive usage`, `5dive digest`), building or editing multi-agent loops — a relay where each step hands off automatically with optional human gates (`task loop start`/`loop ls`) or a maker→verifier review loop (`task add --verifier`, `task reject`, `task loops`), or decomposing an outcome into a guardrailed task DAG (`5dive goal add`) — hiring a ready-made persona off the agent market (`5dive market`, `5dive hire --from-market`) or firing one (`5dive fire`), declarative fleets (`5dive up`, `5dive team import`), hosting a CrewAI crew (`5dive crew`), controlling agents on OTHER registered boxes (`5dive fleet`), running a self-steering objective bound to a live metric (`5dive objective`, `objective replan`), convening a governance vote (`5dive council convene`, `council gate-clear`, `council schedule add` for a recurring convene), the onboarding wizard (`5dive company`), or a delegated GitHub push-for-review (`5dive push`, needs `agent create --can-push`). When a request came over a chat channel (Telegram/Discord `<channel>` tag) and another agent should handle it, pass the chat context via `--reply-to-chat=<id> --reply-to-msg=<id>` so that agent replies from its own bot — don't relay. Always prefer `5dive` over running coding CLIs by hand.
Conversational discovery — adapts from quick scoping (3-5 questions) to deep interviews (multi-round). Talk until we're clear, then build. Produces inline decisions; optionally saves spec.md or scope contract. Not for multi-perspective debate (use agent-room). Not for decomposing work (use task-breakdown).
Trains large language models (2B-462B parameters) using NVIDIA Megatron-Core with advanced parallelism strategies. Use when training models >1B parameters, need maximum GPU efficiency (47% MFU on H100), or require tensor/pipeline/sequence/context/expert parallelism. Production-ready framework used for Nemotron, LLaMA, DeepSeek.
Auto-activates when working with implementation plans. Triggers on "continue the plan", "next task", "what's the plan status", "run task 2.1", or when user references plans/*.plan.md files. Not for creating plans - use /superplan command for that.