doca-flow-tune
Use this skill when the user is tuning a live or captured `doca-flow` pipeline with `doca_flow_tune` — snapshotting pipe / counter / KPI state, picking a tuning axis (rule placement, resource hints / table sizing, HW-offload mode) and a matching measurement (rule-install rate, lookup latency, hardware-counter delta), running offline or online (read-only or state-changing) modes, reading the dumper CSV / analyze JSON / visualize mermaid, or applying a recommendation back into the Flow program. Trigger even when the user does not explicitly mention "doca_flow_tune" — typical implicit phrasings
npx skills add NVIDIA/skills --skill doca-flow-tune --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
What it does
Instructs the agent to use the doca_flow_tune tool to tune a doca-flow pipeline by choosing a three-axis configuration (target Flow pipeline × tuning axis × measurement), selecting offline or online modes or server-attach mode, and initiating a cycle of snapshotting, analyzing, and visualizing outputs, with the option to apply a recommendation back into the Flow program via minimum-diff changes. It also directs how to handle potential questions about when to use this tool versus baselines from doca-flow-perf, and where to route for setup or alternative tasks.
How it works
The skill exposes five top-level subcommands for doca_flow_tune: dump, monitor, web, analyze, visualize. Dump/monitor/web run in server-attached online mode against a live doca-flow via a Unix-domain socket path defined in flow_tune_cfg*.json. Analyze/visualize run in offline mode against JSON/CSV outputs produced by online runs. The tool provides guidance to decide the three-axis configuration, the mode, and the targeted measurement, and to perform a before/after snapshot for a recommended change. It emphasizes using minimum-diff changes and warns about mutating actions after verifying safety via the test loop. It references TASKS.md and CAPABILITIES.md for detailed steps, state interpretations, and safety policies.
When to use it
Load this skill when the user intends to invoke doca_flow_tune on a running or planned doca-flow application to characterize, dump, visualize, analyze, or tune that pipeline. It covers choosing between offline analyze/visualize on captured state, online dumper/monitor against live Flow, or attach-to-app server usage, and selecting which tuning axis and which measurement to compare against a prior baseline. It also instructs on diagnosing cases where output is empty or a visualize step is degenerate. It explicitly cautions routing to other skills for setup, baseline measurements, or code changes beyond minimal-diff recommendations.
What it can touch
The skill describes interacting with the doca_flow_tune binary and its companion analyzer/visualizer scripts and JSON config templates in the system. It notes the dependencies on a DOCA-installed host, a running or captured doca-flow application, and access to flow_tune_cfg*.json templates and the outputs produced by the tool. It mentions the need to read the user's local DOCA install and the provided templates, and to observe outputs from the dumper/monitor/analyze/visualize paths. It does not specify additional external touch points beyond those, and directs to the installed scripts and JSON configurations.
Caveats
The skill notes that it is a thin loader with substantive material in CAPABILITIES.md and TASKS.md, and that mutating recommendations carry high stakes due to live Flow state changes. It emphasizes that if DOCA is not installed, the user should route to doca-setup, and if there is no running doca-flow app, route to doca-flow. It prohibits shipping pre-baked outputs or wrappers and warns against using this tool for baseline measurement methodology or DOCA setup tasks. All statements reflect the tool’s described behavior and routing guidance, with no speculative outcomes.
# DOCA Flow Tune (`doca_flow_tune`) > **Subcommand surface correction (Run-12, verified Run-13 > against doca/tools/flow_tune/src/tune/common/tune_config.cpp).** > `doca_flow_tune` is a single binary whose **role on a given > invocation is determined by which of five top-level > subcommands** the user picks — `dump`, `monitor`, `web`, > `analyze`, `visualize` (case-insensitive on the CLI; > upperc
What does the doca-flow-tune skill do?
Use this skill when the user is tuning a live or captured `doca-flow` pipeline with `doca_flow_tune` — snapshotting pipe / counter / KPI state, picking a tuning axis (rule placement, resource hints / table sizing, HW-offload mode) and a matching measurement (rule-install rate, lookup latency, hardware-counter delta), running offline or online (read-only or state-changing) modes, reading the dumper CSV / analyze JSON / visualize mermaid, or applying a recommendation back into the Flow program. Trigger even when the user does not explicitly mention "doca_flow_tune" — typical implicit phrasings
How do I install it?
Run `npx skills add NVIDIA/skills --skill doca-flow-tune --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From NVIDIA/skills, a repository with 2,789 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.