Loading...
Loading...
Found 13,657 Skills
Orchestrates modular PyTorch profiler trace analysis with TraceLens: generates perf reports, prepares category data, runs system-level and compute-kernel subagents in parallel, validates outputs, and writes a prioritized stakeholder report (analysis.md). Use when the user asks to follow the analysis orchestrator, run the agentic analysis workflow, analyze a trace, compare two traces, or mentions standalone or comparative TraceLens analysis.
Autonomously optimizes end-to-end LLM inference throughput on AMD Instinct GPUs and reports a validated gain, using the Hyperloom multi-agent optimizer. Given a model, framework, workload (TP/EP, concurrency, ISL/OSL, precision), an objective and a time budget, it explores per-workload which levers to pull (serving/config parameters and env, framework enablement and source patches, and hot GPU-kernel rewrites), benchmarks each candidate, and returns the optimization stack that produced the gain. Use when the user wants to make a model serve faster, raise tokens/sec or throughput, optimize or tune vLLM or SGLang on MI300X/MI325X/MI355X, run Hyperloom, run the kernel-agent, quantize-then-optimize with Quark, set up Hyperloom from scratch, or resume a Hyperloom session. Do not use to stand up a server for plain serving, diagnose a broken ROCm install, or run a one-off kernel/benchmark or trace analysis without the optimization loop.
Deep-dive into any codebase and produce structured knowledge artifacts that turn a coding agent into a codebase specialist. This skill should be used when the user asks to "study this repo", "understand this codebase", "document this project", "onboard me onto this code", "create codebase knowledge", "map this architecture", or when asked to produce knowledge artifacts for any agent working on an unfamiliar repository.
UiPath AgentHub MCP server registration + resource-tool authoring via `uip agenthub mcp` (six server types: uipath / coded / command / remote / platform / swagger) and `uip agenthub mcp-tools` on `uipath`-type servers. For Python MCP servers / coded-agent integration→uipath-agents.
Approach board for a problem or a decision. Dispatches independent seats — some generating candidate approaches, including one agent outside this process entirely, some attacking the assembled set comparatively — then verifies the surviving objections and reports a ranked recommendation with its trade-offs. An approach already on the table enters as one candidate among several. Autonomous; changes nothing.
Compact the current conversation into a handoff so a fresh agent, or future-you, can pick up the work cold. With a `doc`, `document`, or `--doc` keyword, writes a per-project handoff file and prints only its path plus a shortcut to start the next session.
Make the agent speak and write in ASD-STE100 Simplified Technical English: short sentences, approved words, active voice, imperative for instructions, one idea per sentence, American spelling. Shapes every reply to the user AND every technical document, rewrite, or STE compliance check. Invoke with /speak-ste; stays on until "stop ste mode".
Use when user says "create workflow", "create a workflow", "design workflow", "orchestrate", "automate multiple steps", "coordinate agents", "multi-agent workflow". Creates orchestration workflows from natural language using Socratic questioning to plan multi-agent workflows with visualization.
Extended `5dive` CLI recipes beyond the everyday core — see the `5dive-cli` skill first for spawning/messaging sibling agents and the basic task queue. Use THIS skill for hiring a ready-made persona off the agent market (`5dive market`, `hire --from-market`) or firing one (`5dive fire`), auth recovery (`error.class=auth_required`, `--defer-auth`, device-code login via `agent auth start/poll/submit`), BYO-provider agents (`--provider`), multi-account auth (`5dive account`), declarative fleets and company templates (`5dive up/down/ps/export`, `team import`), hosting a CrewAI crew (`5dive crew`), controlling agents on OTHER registered boxes (`5dive fleet`), recurring/scheduled work (`task add --recurring`, `5dive heartbeat`) and projects (`5dive project add`), building or editing multi-agent loops — a relay with optional human gates (`task loop start`) or a maker→verifier review loop (`task add --verifier`, `task reject`, `5dive loop` LOOP-7 verbs) — decomposing an outcome into a guardrailed task DAG (`5dive goal add`) or a self-steering objective bound to a live metric (`5dive objective`), compiling durable knowledge into the shared wiki (`5dive memory add`), org-chart writes (`5dive org set`), convening a governance vote (`5dive council`), reading fleet health / token burn / the daily standup (`5dive supervisor`, `5dive usage`, `5dive digest`), a machine-readable health check (`5dive doctor --json`, `5dive selfcheck --json`), a task's causal history (`5dive trace`), the current model id per alias (`5dive models`), Telegram/Discord pairing and shared team-bot setup, a delegated GitHub push-for-review (`5dive push`), or the onboarding wizard (`5dive company`).
Spawn parallel adversarial subagents that DRIVE THE REAL SYSTEM (not reason from source) to break new code before it ships — schemas, contracts, parsers, serializers, protocols, state machines, security/redaction, freeze guards. Use as a standard step on any nontrivial feature/task once the implementation is green and BEFORE the final convergence (dootsabha) review. Each agent gets one disjoint attack surface + a "try to break X" charter; every finding is independently reproduced before it is believed, then fixed with a regression test. Trigger phrases include "adversarial review", "try to break this", "find the bugs I missed", "harden before done", "attack this schema/parser/contract", and the shux feature-protocol adversarial step.
Orchestrate parallel subtasks across multiple repositories using vibe-kanban MCP. Use this skill when: (1) Work needs to be distributed across multiple repos, (2) Parallel execution of independent tasks is needed, (3) You need to create subtasks and collect their results, (4) Multi-agent coordination via kanban task descriptions. Triggers: "subtask", "multi-repo", "parallel tasks", "distribute work", "orchestrate".
Executing a phased plan in rolling waves where each phase runs as one multi-agent workflow harness: the supervisor elaborates the phase into tasks against the real landed code, launches a workflow that implements with TDD and runs mandatory in-harness review plus an adversarial contrarian pass (and researchers when the phase hits an unknown) before returning verified work, then reviews it, checkpoints with the user, and rolls to the next phase. Use when each wave should be a reviewed multi-agent harness, not a lone subagent. Skip when one supervised subagent per wave suffices (ring:executing-plans) or the full gated cycle is wanted (ring:running-dev-cycle).