DOCA Flow Tune (doca_flow_tune)
Examples
- "Flow rule-install rate is low on BlueField." — this is
in scope. Capture a monitor-mode baseline, then pick one
tuning axis.
- "Tune visualize step is empty." — check the dumper JSON/CSV /
analyze JSON before changing the Flow program.
Instructions
Follow Loading order below. Open the tasks companion for configure /
tune / debug. Open the capabilities companion for online vs
offline modes. Do not open every companion file.
Command surface (DOCA 3.5 public Flow Tune guide).
doca_flow_tune is a single binary whose role on a given
invocation is determined by which of three top-level
commands the user picks — monitor, analyze,
visualize (confirm with --help on the installed
version); use only the commands the public guide and the
installed --help list.
Counter dumping is part of monitor (--enable-dump, or the
background sub-command for dumping without screen output),
not a separate command; analyze accepts the export and
packet-trace sub-commands (hyphenated). monitor and
analyze run the binary in server-attached online mode
against a live doca-flow application reached over a Unix-
domain socket whose path lives in network.server_uds of the
JSON configuration file; visualize (and offline
analysis of a previously exported pipeline file) runs in
offline / captured-snapshot mode against JSON / CSV files
the online commands previously dropped into the
configured outputs_directory. The rest of this skill (and
the capabilities companion / the tasks companion)
uses the legacy "server role / online mode / offline mode"
framing — that framing is internally consistent with the
command surface here: server role = a server-attached
online command (monitor, analyze export / analyze packet-trace); online mode = any of those; offline mode =
visualize over captured files. Treat the command name as the
primary handle; treat server/online/offline as the
downstream behavioral consequence of the subcommand pick.
Where to start: This is a tool skill for invoking doca_flow_tune,
the unified DOCA Flow tuning tool. Open the tasks companion and
start at ## configure to commit to the
three-axis decision (target Flow pipeline × tuning axis ×
measurement) and pick offline vs online vs server-attach mode, then
## run for the snapshot → analyze → visualize
loop, then ## test for the smoke-before-bulk
overlay that gates any state-changing application of a tuning
recommendation back into the Flow application's code. Open
the capabilities companion when the question is what
state doca_flow_tune can observe and recommend on, how its
server / client roles fit inside the single artifact, which DOCA
version the tool ships in, or how to interpret the dumper / monitor
/ analyze / visualize outputs without fooling yourself. If DOCA is
not installed, route to
doca-setup/SKILL.md first; if the user has
no running doca-flow application yet, route to
doca-flow/SKILL.md — flow-tune does not
create pipes, it observes and recommends on top of pipes the
library already created.
Example questions this skill answers well
The CLASSES of doca_flow_tune questions this skill is built to
answer, each with one worked example. The class is the load-bearing
piece; the worked example is one instance.
- "Do I measure first or tune first?" — worked example: "my
doca-flow service runs on a BlueField-3 and I think the
rule-install rate is below what the device can sustain".
Answered by the measure, then tune rule in
the capabilities companion: a monitor-mode baseline, one
tuning axis, then a monitor-mode re-capture on the same axis.
- "Capture a snapshot of a live
doca-flow pipeline's hardware
and software counters without touching the dataplane." — worked
example: "I want a side-effect-free dumper / monitor run against
the running Flow ports for an operations-rate profile". Answered
by the snapshot flow in
the tasks companion plus the read-only-by-default
posture in
the capabilities companion.
- "Pick the right tuning axis — rule placement, resource hints,
or hardware-offload mode — for the question I actually have."
— worked example: "my Flow pipe's rule-install rate is low; is
this a placement question or a table-sizing question?". Answered
by the three-axis configuration in
the capabilities companion
- the configure walk in
the tasks companion.
- "How do
doca_flow_tune's server role and client / consumer
role fit together inside the single artifact?" — worked
example: "I keep reading about a Flow Tune server and a Flow
Tune client; which binary am I running?". Answered by the
one binary, two roles breakdown in
the capabilities companion
and the corresponding routing in
the tasks companion.
- "How do I take a recommended parameter change from flow-tune
back into my doca-flow application without breaking the
dataplane?" — worked example: "the analyze step suggests a
different table sizing for my pipe; how do I apply it?".
Answered by the recommendation → minimum-diff modification of
the Flow program loop in
the tasks companion and the
smoke-before-bulk rule in
the tasks companion.
- "
doca_flow_tune reports nothing / disagrees with the Flow
app / cannot attach — what does that mean?" — worked example:
"the tool runs but the visualize step produces an empty
mermaid diagram". Answered by the layered error taxonomy in
the capabilities companion
Audience
This skill serves external operators, performance engineers,
DOCA Flow application developers, and AI agents who need to
understand, characterize, or improve a running doca-flow
pipeline's behavior on the user's actual install and device.
Concretely:
- An operator wanting a read-only snapshot of which pipes exist
and how their hardware / software counters progress before
recommending any change.
- A performance engineer who has a measured baseline (for
example a monitor-mode capture) and
wants to turn measurement into optimization — pick a tuning
axis and identify which knob in the doca-flow program is the
lever.
- A Flow developer using the offline analyze + visualize loop to
understand a pipe layout without re-instrumenting the program.
- An AI agent driving the "is this pipeline behaving as expected,
and would a non-mutating tuning hint help" triage step.
It is not for users debugging the doca_flow_tune source code,
not a substitute for the live public DOCA Flow Tune guide,
not the place to learn the doca-flow API (see
doca-flow/SKILL.md), and not a
synthetic rule-rate benchmark: it measures the user's own
pipeline.
doca_flow_tune is shipped as a single tool (one binary plus
companion analyzer / visualizer scripts and JSON config
templates) — the historical server and client roles live
inside this one artifact, not in two executables. It uses the
same kind: tool three-file shape as the rest of the bundle.
Language scope
This skill governs invocation, output interpretation, and
recommendation-to-code-change routing for the DOCA Flow
application doca_flow_tune observes. The tool is not a
programming target — there is no public API to link against; the
workflow is configure JSON, run, read outputs, propose minimum-
diff changes to the surrounding doca-flow program. For the
doca-flow API the recommendations route back into, see
doca-flow/SKILL.md;
for cross-language application patterns, see
doca-programming-guide/SKILL.md.
When to load this skill
Load this skill when the user is — or the agent needs to — invoke
doca_flow_tune against a running or planned doca-flow
application to characterize, dump, visualize, analyze, or tune
that pipeline. Concretely:
- Picking which role to engage (offline analyze / visualize on
a captured config + state, online dumper / monitor against the
live app, or attach-to-app server-role usage).
- Picking which tuning axis to ask about (rule placement,
resource hints / table sizing, hardware-offload-mode).
- Picking which measurement axis to compare against
(rule-install rate, lookup latency, hardware-counter delta) —
not interchangeable; use the axis the baseline
capture named.
- Capturing a documented before / after pair around a proposed
Flow-program change (JSON config path, command line, DOCA
version, device, environment, unredacted tool output).
- Diagnosing an empty session, a degenerate visualize diagram, or
an analyze recommendation that contradicts live counters.
Do not load this skill for general DOCA orientation, Flow
program API work, install, or pure measurement methodology —
route to doca-public-knowledge-map/SKILL.md,
doca-flow/SKILL.md, or
doca-setup/SKILL.md.
What this skill provides
This is a thin loader. Substantive material lives in two
companion files:
- the capabilities companion — what
doca_flow_tune observes and
recommends on: the unified-artifact decomposition (server role
- client / consumer role inside one binary), the three-axis
configuration model (tuning axis × measurement × scope: which
pipe / port / app), the documented offline / online / attach
modes, the JSON configuration-file shape (the
flow_tune_cfg.json
example in the public Flow Tune guide), the dumper / monitor / analyze /
visualize output surfaces, the version overlay (this tool rides
the doca-flow library version it observes; the canonical
rules live in doca-version/SKILL.md),
the layered error taxonomy
(config-syntax / attach-failed / pipe-not-found /
measurement-unsound / recommendation-unactionable / version /
cross-cutting), the observability posture (the tool is an
observability primitive for the Flow pipeline), and the safety
policy that makes any mutating application of a recommendation
high-stakes because the recommendation lands in live Flow state.
- the tasks companion — step-by-step workflows for the in-scope task verbs:
install (route to setup; the binary is shipped),
configure (the three-axis decision + JSON config + mode pick),
build (route to install; the binary is shipped), modify
(apply a recommendation back to the Flow program via minimum-
diff), run (the snapshot → analyze → visualize flow), test
(the eval loop — warm-up, steady-state, before / after pair,
client / server / Flow version match), debug (walk the error
taxonomy layer by layer), use (the agent-side workflow for
consuming flow-tune output), plus a Deferred task verbs block
and a Command appendix.
The skill assumes a host where DOCA is already installed (or the
public NGC DOCA container is running) and a doca-flow
application is already created and validated per the
doca-flow/SKILL.md skill. Without those
preconditions, the tune session has nothing to observe.
What this skill deliberately does not ship
This skill is agent guidance, not a samples or scripts bundle.
To keep the boundary clean, it deliberately does not contain —
and pull requests should not add:
- Verbatim flag inventories, subcommand names, JSON config field
names, or default endpoint paths quoted as the contract. The
public DOCA Flow Tune guide (reached via
doca-public-knowledge-map ## DOCA tools (doca-public-knowledge-map/SKILL.md#doca-tools))
and the installed --help are the joint source of truth; the
guide's configuration-file example is the second source for
the JSON schema. Copying them here rots when the tool evolves.
- Pre-baked example output (dumper CSV columns, analyzer JSON
field names, visualizer mermaid output). Output is install-,
device-, firmware-, and version-specific; a captured example
pinned to one platform misleads operators elsewhere.
- Wrappers, parsers, or scripts that consume flow-tune output.
The right answer is to read the tool's own dump, analyze and
visualize output files directly and quote them.
- Pre-baked tuning recommendations. Recommendations are
install-, device-, firmware-, and workload-specific; the agent
always re-derives them from the user's actual session.
- A
samples/, templates/, or reference/ subtree. Mock
tuning recipes read as production-grade and mislead.
Loading order
- Read this skill first to confirm the user's question is
in scope (the user actually wants to invoke
doca_flow_tune
against a doca-flow pipeline, not measure baseline perf or
learn the Flow API).
- For what
doca_flow_tune observes, the one-binary / two-role
decomposition, the three-axis model, the version overlay, the
error taxonomy, observability surface, and safety posture,
see the capabilities companion.
- For the documented invocations and the snapshot → analyze →
visualize → propose → smoke workflow —
install, configure,
build, modify, run, test, debug, use — see
the tasks companion.
Related skills
doca-flow/SKILL.md — the base
library whose pipeline this tool observes and tunes. The
pipe / entry / rule surface flow-tune reports on is created by
doca-flow program code; recommendations route back into that
program via the universal modify-a-sample workflow.
doca-public-knowledge-map/SKILL.md
— routing to the public DOCA Flow Tune page on docs.nvidia.com
and the rest of the public DOCA documentation set.
doca-version/SKILL.md — the canonical
version-detection chain, four-way match, NGC semantics, and
headers-win-over-docs rule. The
## Version compatibility
overlay in this skill is a thin extension on top.
doca-debug/SKILL.md — the cross-cutting
debug ladder. Flow-tune surfaces its own error taxonomy; when
the cause turns out to be below DOCA (driver, firmware, NUMA),
the tune taxonomy hands off to doca-debug.
doca-structured-tools-contract/SKILL.md
— the bundle's detect → prefer → fall back → report contract.
The Command appendix in the tasks companion honors it.
doca-setup/SKILL.md — env preparation,
install verification, hugepages, NUMA, and the I have no
install yet path with the public NGC DOCA container.
doca-hardware-safety/SKILL.md —
the cross-cutting hardware-safety meta-policy this skill's
## Safety policy overlays.
1---2license: Apache-2.03name: doca-flow-tune4description: >5 Use this skill when the user is tuning a live or captured `doca-flow`6 pipeline with `doca_flow_tune` — snapshotting pipe / counter / KPI7 state, picking a tuning axis (rule placement, resource hints / table8 sizing, HW-offload mode) and a matching measurement (rule-install9 rate, lookup latency, hardware-counter delta), running offline or10 online (read-only or state-changing) modes, reading the dumper CSV /11 analyze JSON / visualize mermaid, or applying a recommendation back12 into the Flow program. Trigger even when the user does not explicitly13 mention "doca_flow_tune" — typical implicit phrasings include "Flow14 rule-install rate is low on BlueField", "table sizing looks wrong for15 this pipe", "tune visualize step is empty", "before/after counters16 don't move", or "which doca-flow knob does this recommendation hit".17 Refuse and route elsewhere for writing the doca-flow18 application, DOCA install, or streaming Flow telemetry — those belong19 to other skills.20metadata:21 source: https://skillmd.com/skills/nvidia/doca-flow-tune22 author: NVIDIA <doca-feedback@nvidia.com>23 kind: tool24 doca_version: "3.5"25 tags:26 - doca27 - flow28 - tune29compatibility: >30 Requires DOCA SDK installed at /opt/mellanox/doca on Linux (Ubuntu31 22.04/24.04 or RHEL/SLES) with a BlueField DPU or ConnectX NIC32 attached, plus a running or captured `doca-flow` application to33 observe. Reads the user's local install via `pkg-config doca-flow` and34 the installed `/opt/mellanox/doca/tools/doca_flow_tune`. The validation steps require a live DOCA35 install.36---3738# DOCA Flow Tune (`doca_flow_tune`)3940## Examples4142- **"Flow rule-install rate is low on BlueField."** — this is43 in scope. Capture a monitor-mode baseline, then pick one44 tuning axis.45- **"Tune visualize step is empty."** — check the dumper JSON/CSV /46 analyze JSON before changing the Flow program.4748## Instructions4950Follow Loading order below. Open the tasks companion for configure /51tune / debug. Open the capabilities companion for online vs52offline modes. Do not open every companion file.5354> **Command surface (DOCA 3.5 public Flow Tune guide).**55> `doca_flow_tune` is a single binary whose **role on a given56> invocation is determined by which of three top-level57> commands** the user picks — `monitor`, `analyze`,58> `visualize` (confirm with `--help` on the installed59> version); use only the commands the public guide and the60> installed `--help` list.61> Counter dumping is part of `monitor` (`--enable-dump`, or the62> `background` sub-command for dumping without screen output),63> not a separate command; `analyze` accepts the `export` and64> `packet-trace` sub-commands (hyphenated). `monitor` and65> `analyze` run the binary in **server-attached online mode**66> against a live `doca-flow` application reached over a Unix-67> domain socket whose path lives in `network.server_uds` of the68> JSON configuration file; `visualize` (and offline69> analysis of a previously exported pipeline file) runs in70> **offline / captured-snapshot mode** against JSON / CSV files71> the online commands previously dropped into the72> configured `outputs_directory`. The rest of this skill (and73> the capabilities companion / the tasks companion)74> uses the legacy *"server role / online mode / offline mode"*75> framing — that framing is internally consistent with the76> command surface here: *server role* = a server-attached77> online command (`monitor`, `analyze export` / `analyze78> packet-trace`); *online mode* = any of those; *offline mode* =79> `visualize` over captured files. Treat the command name as the80> primary handle; treat *server/online/offline* as the81> downstream behavioral consequence of the subcommand pick.8283**Where to start:** This is a tool skill for invoking `doca_flow_tune`,84the unified DOCA Flow tuning tool. Open the tasks companion and85start at `## configure` to commit to the86three-axis decision (target Flow pipeline × tuning axis ×87measurement) and pick offline vs online vs server-attach mode, then88`## run` for the snapshot → analyze → visualize89loop, then `## test` for the smoke-before-bulk90overlay that gates any state-changing application of a tuning91recommendation back into the Flow application's code. Open92the capabilities companion when the question is *what93state `doca_flow_tune` can observe and recommend on*, *how its94server / client roles fit inside the single artifact*, *which DOCA95version the tool ships in*, or *how to interpret the dumper / monitor96/ analyze / visualize outputs without fooling yourself*. If DOCA is97not installed, route to98`doca-setup/SKILL.md` first; if the user has99no running `doca-flow` application yet, route to100`doca-flow/SKILL.md` — flow-tune does not101create pipes, it observes and recommends on top of pipes the102library already created.103104## Example questions this skill answers well105106The CLASSES of `doca_flow_tune` questions this skill is built to107answer, each with one worked example. The class is the load-bearing108piece; the worked example is one instance.109110- **"Do I measure first or tune first?"** — worked example: *"my111 doca-flow service runs on a BlueField-3 and I think the112 rule-install rate is below what the device can sustain"*.113 Answered by the *measure, then tune* rule in114 the capabilities companion: a monitor-mode baseline, one115 tuning axis, then a monitor-mode re-capture on the same axis.116- **"Capture a snapshot of a live `doca-flow` pipeline's hardware117 and software counters without touching the dataplane."** — worked118 example: *"I want a side-effect-free dumper / monitor run against119 the running Flow ports for an operations-rate profile"*. Answered120 by the snapshot flow in121 the tasks companion plus the read-only-by-default122 posture in123 the capabilities companion.124- **"Pick the right tuning axis — rule placement, resource hints,125 or hardware-offload mode — for the question I actually have."**126 — worked example: *"my Flow pipe's rule-install rate is low; is127 this a placement question or a table-sizing question?"*. Answered128 by the three-axis configuration in129 the capabilities companion130 + the configure walk in131 the tasks companion.132- **"How do `doca_flow_tune`'s server role and client / consumer133 role fit together inside the single artifact?"** — worked134 example: *"I keep reading about a Flow Tune server and a Flow135 Tune client; which binary am I running?"*. Answered by the136 *one binary, two roles* breakdown in137 the capabilities companion138 and the corresponding routing in139 the tasks companion.140- **"How do I take a recommended parameter change from flow-tune141 back into my doca-flow application without breaking the142 dataplane?"** — worked example: *"the analyze step suggests a143 different table sizing for my pipe; how do I apply it?"*.144 Answered by the *recommendation → minimum-diff modification of145 the Flow program* loop in146 the tasks companion and the147 smoke-before-bulk rule in148 the tasks companion.149- **"`doca_flow_tune` reports nothing / disagrees with the Flow150 app / cannot attach — what does that mean?"** — worked example:151 *"the tool runs but the visualize step produces an empty152 mermaid diagram"*. Answered by the layered error taxonomy in153 the capabilities companion154 + the tasks companion.155156## Audience157158This skill serves **external operators, performance engineers,159DOCA Flow application developers, and AI agents who need to160understand, characterize, or improve a running `doca-flow`161pipeline's behavior on the user's actual install and device**.162Concretely:163164- An operator wanting a read-only snapshot of which pipes exist165 and how their hardware / software counters progress before166 recommending any change.167- A performance engineer who has a measured baseline (for168 example a monitor-mode capture) and169 wants to turn *measurement* into *optimization* — pick a tuning170 axis and identify which knob in the doca-flow program is the171 lever.172- A Flow developer using the offline analyze + visualize loop to173 understand a pipe layout without re-instrumenting the program.174- An AI agent driving the *"is this pipeline behaving as expected,175 and would a non-mutating tuning hint help"* triage step.176177It is **not** for users debugging the `doca_flow_tune` source code,178**not** a substitute for the live public DOCA Flow Tune guide,179**not** the place to learn the `doca-flow` API (see180`doca-flow/SKILL.md`), and **not** a181synthetic rule-rate benchmark: it measures the user's own182pipeline.183184`doca_flow_tune` is shipped as a **single tool** (one binary plus185companion analyzer / visualizer scripts and JSON config186templates) — the historical *server* and *client* roles live187inside this one artifact, not in two executables. It uses the188same `kind: tool` three-file shape as the rest of the bundle.189190## Language scope191192This skill governs invocation, output interpretation, and193recommendation-to-code-change routing for the DOCA Flow194application `doca_flow_tune` observes. The tool is not a195programming target — there is no public API to link against; the196workflow is *configure JSON, run, read outputs, propose minimum-197diff changes to the surrounding `doca-flow` program*. For the198`doca-flow` API the recommendations route back into, see199`doca-flow/SKILL.md`;200for cross-language application patterns, see201`doca-programming-guide/SKILL.md`.202203## When to load this skill204205Load this skill when the user is — or the agent needs to — invoke206`doca_flow_tune` against a running or planned `doca-flow`207application to characterize, dump, visualize, analyze, or tune208that pipeline. Concretely:209210- Picking *which* role to engage (offline analyze / visualize on211 a captured config + state, online dumper / monitor against the212 live app, or attach-to-app server-role usage).213- Picking *which* tuning axis to ask about (rule placement,214 resource hints / table sizing, hardware-offload-mode).215- Picking *which* measurement axis to compare against216 (rule-install rate, lookup latency, hardware-counter delta) —217 not interchangeable; use the axis the baseline218 capture named.219- Capturing a documented before / after pair around a proposed220 Flow-program change (JSON config path, command line, DOCA221 version, device, environment, unredacted tool output).222- Diagnosing an empty session, a degenerate visualize diagram, or223 an analyze recommendation that contradicts live counters.224225Do **not** load this skill for general DOCA orientation, Flow226program API work, install, or pure measurement methodology —227route to `doca-public-knowledge-map/SKILL.md`,228`doca-flow/SKILL.md`, or229`doca-setup/SKILL.md`.230231## What this skill provides232233This is a **thin loader**. Substantive material lives in two234companion files:235236- the capabilities companion — what `doca_flow_tune` observes and237 recommends on: the unified-artifact decomposition (server role238 + client / consumer role inside one binary), the three-axis239 configuration model (tuning axis × measurement × scope: which240 pipe / port / app), the documented offline / online / attach241 modes, the JSON configuration-file shape (the `flow_tune_cfg.json`242 example in the public Flow Tune guide), the dumper / monitor / analyze /243 visualize output surfaces, the version overlay (this tool rides244 the `doca-flow` library version it observes; the canonical245 rules live in `doca-version/SKILL.md`),246 the layered error taxonomy247 (config-syntax / attach-failed / pipe-not-found /248 measurement-unsound / recommendation-unactionable / version /249 cross-cutting), the observability posture (the tool *is* an250 observability primitive for the Flow pipeline), and the safety251 policy that makes any mutating application of a recommendation252 high-stakes because the recommendation lands in live Flow state.253- the tasks companion — step-by-step workflows for the in-scope task verbs:254 `install` (route to setup; the binary is shipped),255 `configure` (the three-axis decision + JSON config + mode pick),256 `build` (route to install; the binary is shipped), `modify`257 (apply a recommendation back to the Flow program via minimum-258 diff), `run` (the snapshot → analyze → visualize flow), `test`259 (the eval loop — warm-up, steady-state, before / after pair,260 client / server / Flow version match), `debug` (walk the error261 taxonomy layer by layer), `use` (the agent-side workflow for262 consuming flow-tune output), plus a `Deferred task verbs` block263 and a `Command appendix`.264265The skill assumes a host where DOCA is already installed (or the266public NGC DOCA container is running) and a `doca-flow`267application is already created and validated per the268`doca-flow/SKILL.md` skill. Without those269preconditions, the tune session has nothing to observe.270271## What this skill deliberately does not ship272273This skill is **agent guidance**, not a samples or scripts bundle.274To keep the boundary clean, it deliberately does not contain —275and pull requests should not add:276277- **Verbatim flag inventories, subcommand names, JSON config field278 names, or default endpoint paths quoted as the contract.** The279 public DOCA Flow Tune guide (reached via280 `doca-public-knowledge-map ## DOCA tools` (`doca-public-knowledge-map/SKILL.md#doca-tools`))281 and the installed `--help` are the joint source of truth; the282 guide's configuration-file example is the second source for283 the JSON schema. Copying them here rots when the tool evolves.284- **Pre-baked example output (dumper CSV columns, analyzer JSON285 field names, visualizer mermaid output).** Output is install-,286 device-, firmware-, and version-specific; a captured example287 pinned to one platform misleads operators elsewhere.288- **Wrappers, parsers, or scripts** that consume flow-tune output.289 The right answer is to read the tool's own dump, analyze and290 visualize output files directly and quote them.291- **Pre-baked tuning recommendations.** Recommendations are292 install-, device-, firmware-, and workload-specific; the agent293 always re-derives them from the user's actual session.294- **A `samples/`, `templates/`, or `reference/` subtree.** Mock295 tuning recipes read as production-grade and mislead.296297## Loading order2982991. Read this skill first to confirm the user's question is300 in scope (the user actually wants to invoke `doca_flow_tune`301 against a `doca-flow` pipeline, not measure baseline perf or302 learn the Flow API).3032. **For what `doca_flow_tune` observes, the one-binary / two-role304 decomposition, the three-axis model, the version overlay, the305 error taxonomy, observability surface, and safety posture,306 see the capabilities companion.**3073. **For the documented invocations and the snapshot → analyze →308 visualize → propose → smoke workflow — `install`, `configure`,309 `build`, `modify`, `run`, `test`, `debug`, `use` — see310 the tasks companion.**311312## Related skills313314- `doca-flow/SKILL.md` — the **base315 library** whose pipeline this tool observes and tunes. The316 pipe / entry / rule surface flow-tune reports on is created by317 `doca-flow` program code; recommendations route back into that318 program via the universal modify-a-sample workflow.319- `doca-public-knowledge-map/SKILL.md`320 — routing to the public DOCA Flow Tune page on `docs.nvidia.com`321 and the rest of the public DOCA documentation set.322- `doca-version/SKILL.md` — the canonical323 version-detection chain, four-way match, NGC semantics, and324 headers-win-over-docs rule. The325 `## Version compatibility`326 overlay in this skill is a thin extension on top.327- `doca-debug/SKILL.md` — the cross-cutting328 debug ladder. Flow-tune surfaces *its own* error taxonomy; when329 the cause turns out to be below DOCA (driver, firmware, NUMA),330 the tune taxonomy hands off to `doca-debug`.331- `doca-structured-tools-contract/SKILL.md`332 — the bundle's detect → prefer → fall back → report contract.333 The Command appendix in the tasks companion honors it.334- `doca-setup/SKILL.md` — env preparation,335 install verification, hugepages, NUMA, and the *I have no336 install yet* path with the public NGC DOCA container.337- `doca-hardware-safety/SKILL.md` —338 the cross-cutting hardware-safety meta-policy this skill's339 `## Safety policy` overlays.