DOCA setup
Where to start: If the user's question is deployment-shaped
("how do I deploy", "how do I run my DOCA workload", "I just got a
BlueField, what now", "my code is built, what next"), walk
TASKS.md ## recognize first. It is the
bundle's front-door: it detects the system shape (host x86 / BlueField
Arm bare-metal / DPU-only / fresh laptop), asks the developer the
minimal set of questions needed to disambiguate, and routes to the
correct downstream skill — the container deployment path
(doca-container-deployment/SKILL.md),
the bare-metal hardware deployment path
(doca-bare-metal-deployment/SKILL.md),
or the no-hardware fallback
(TASKS.md ## no-install). The wrong failure
mode is to silently steer every developer onto containers because the
agent loaded that skill first; ## recognize exists to prevent that.
If the user does not have DOCA installed yet and the request is not
deployment-shaped, jump straight to
TASKS.md ## no-install for the NGC container
path. Deployment-shaped requests still enter ## recognize first,
which routes fresh-laptop cases to ## no-install. Otherwise read
## When to load this skill to confirm the
question is env-class, then route to the section below that matches the
user's intent.
Do not use this skill for:
- Defining, validating or writing packet-steering rules (DOCA Flow pipes,
matches, actions, entries), including when the user says the environment
and versions are already verified. That belongs to
doca-flow: name it
and stop. Do not run ## recognize, environment checks or the
verification contract for it.
- Other library API work (RDMA queues, Comch channels, ...): the matching
library skill.
- First-app derivation from a sample, build structure, or what a
DOCA_ERROR_* means: doca-programming-guide.
- "Where is X documented":
doca-public-knowledge-map.
Examples
The CLASSES this skill is built to handle, each with one worked example.
The skill must answer the class; the worked example is illustrative.
Instructions
Follow ## Loading order. Open TASKS.md or
CAPABILITIES.md as named there. Do not open every companion file.
- "I want to deploy a DOCA workload — what is the right path?" —
worked example: "I just got a BlueField; my code is built; what
now?" Resolved by the front-door decision tree in
TASKS.md ## recognize, which detects the
system shape, asks the minimum residual question, and routes to
either the container or the bare-metal deployment skill.
- "Verify DOCA is installed and healthy." — worked example:
"Is DOCA Flow actually available on this box?" Resolved by the
install-health snapshot in
TASKS.md ## test plus the
version-detection rules in
CAPABILITIES.md ## Capabilities and modes.
- "I do not have DOCA installed — what now?" — worked example: "I'm
on macOS and want to learn DOCA before I get a BlueField." Resolved
by
TASKS.md ## no-install (NGC DOCA container
as universal Stage-1).
- "Prepare the build environment for any DOCA library." — worked
example: "
pkg-config --cflags doca-flow returns nothing — what's
missing?" Resolved by the build-prep workflow in
TASKS.md ## configure and the build-class
error taxonomy in
CAPABILITIES.md ## Error taxonomy.
- "Prepare the runtime preconditions on a real DPU box." — worked
example: "hugepages / representors / devlink — what's the minimum set
before I run my first DOCA Flow program?" Resolved by
TASKS.md ## configure and the runtime
observability rules in
CAPABILITIES.md ## Observability.
- "Diagnose an env-class failure (install / build / runtime)." —
worked example: "My DOCA Flow program built fine but says
pkg-config cannot find it at runtime." Resolved by the layered
env-class debug workflow in TASKS.md ## debug.
- "Change something about the environment safely." — worked
example: "I want to switch eswitch mode from legacy to switchdev."
Resolved by the safety constraints in
CAPABILITIES.md ## Safety policy.
Hugepage reservation, mlxconfig reset, and eswitch mode changes
are host-global: they can disrupt live networking and representors,
may require a reboot, and can be hard to undo. Obtain explicit user
confirmation before recommending or executing them.
If the question is library-API-shaped (Flow pipe construction, RDMA
queue setup, …) or program-shaped (how to build, modify a sample, debug
the program itself), route to
doca-programming-guide/SKILL.md or the
matching library skill instead — env-class only lives here.
Route the code and stop — but keep the orientation. For a
library-API or program-shaped request, naming the owning skill and
then answering it anyway is the failure mode to avoid: no
pipe-construction steps, no queue-setup sketch, no example code, not
even as a summary. This holds even when the user has already ruled
the environment out ("versions are verified, I just need to define
the rules") — that phrasing removes the env-class hook that would
make it this skill's work.
It is scoped to that work, and stops there. Saying where a
component runs — which side of the BlueField / host split owns it,
what the host talks to it over — is deployment orientation, which
this skill owns as the front door and must still answer in full.
The shared Core lifecycle is the same kind of orientation: name
create → configure (properties, permissions, and tasks before
start) → start → connect → use → stop → destroy, and that
configuration MUST happen before doca_ctx_start(). Then hand the
per-library calls to the owning skill. Handing off the API surface
is not a reason to go quiet about the system shape around the
object.
When to load this skill
Load this skill when the user is dealing with the environment around DOCA — installing it, verifying the install is healthy, preparing the build / runtime preconditions, debugging env-class failures, figuring out how to reach an install from a host that doesn't have one yet, or asking a deployment-shaped question that hasn't yet been routed (containers vs. bare-metal) — the front-door routing decision lives here. Concretely:
- The deployment-shape routing front door: "I just got a BlueField, what now?", "my code is built, how do I run it?", "how do I deploy this?" All of these load
## recognize first so the agent does not silently push the user onto the wrong path.
- Verifying that the DOCA install is healthy and that the build environment can find it (
pkg-config, headers, library paths).
- Preparing the runtime: hugepages,
devlink device visibility, representor enumeration, kernel-module prerequisites.
- Diagnosing common setup-class failures: missing
*.pc file, hugepages not mounted, representor not visible, header-vs-runtime version mismatch.
- The I have no install yet path: the user is on macOS, Windows, or a Linux box without DOCA, and needs to reach an environment where DOCA is actually installed. The canonical Stage-1 answer is the public NGC DOCA container on the catalog Tags page (macOS, Windows, or Linux with Docker Engine or Docker Desktop; no NVIDIA hardware required for the build / read / learn loop). See
TASKS.md ## no-install.
Do not load this skill for:
- "What is DOCA?", "where is the developer guide?", "where is the install layout documented?" — those are routing questions; use
doca-public-knowledge-map/SKILL.md.
- "How do I derive a custom first application from a sample?", "how do I structure a DOCA build?", "what does
DOCA_ERROR_BAD_STATE mean?" — those are programming-class questions and live in doca-programming-guide/SKILL.md, which owns the universal ## modify (first-app derivation), the canonical ## build pattern, the universal lifecycle, and the cross-library DOCA_ERROR_* taxonomy.
- Library-internal API questions (Flow pipe construction, RDMA queue setup, etc.) — those belong in the matching library skill (e.g.
doca-flow/SKILL.md). This skill stops at "the install is healthy and the env is ready"; it does not own program semantics.
What this skill provides
This is a thin loader. The body keeps only the orientation needed to pick the right next file. The substantive env material lives in two companion files:
CAPABILITIES.md — what the install / build / runtime environment surface looks like: install profiles (doca-all, doca-ofed, doca-networking), where the build flavors (release vs trace) live on disk and how to point LD_LIBRARY_PATH at them, the env-side version-detection rules, the env-class error taxonomy (pkg-config not finding doca-flow, hugepages not reserved, representors not visible), what a healthy install looks like under observation, and the safety constraints on environment changes (hugepages global, mlxconfig reset, eswitch mode change).
TASKS.md — env workflows: recognize (the front-door system-shape detect + dev-Q decision tree that routes deployment-shaped questions to either container or bare-metal), configure (env prep), test (install health snapshot), debug (env-class layered diagnosis), and no-install (the I have no install yet procedure with the NGC container as Path 0). Three other anchors (build, modify, run) route to doca-programming-guide/SKILL.md, which owns those verbs after the env / program split.
This skill assumes nothing about whether DOCA is installed — the ## no-install workflow exists precisely for the fresh laptop case.
Loading order
- Classify the question as programming,
knowledge-map, beginner orientation, deployment routing, or
environment work. Programming and knowledge-map questions route to
their owning skill and stop here.
- For beginner orientation, show the
TASKS.md ## no-install Stage 1 vs Stage 2 roadmap
before any greeting, recommendation, or command. The opening must
explicitly name Stage 1 — container learning and Stage 2 —
hardware runtime; do not open by jumping directly to Docker or
installation. Then route Stage 1/container learning to
## no-install Path 0 and Stage 2/hardware runtime to
## no-install Paths A/C, followed by ## recognize before
selecting container versus bare-metal. Ask one clarifying question
when the stage is unknown; if it remains unknown or no reply is
available in unattended execution, stop with
confirmation_required and both paths explained rather than
guessing. Include the authoritative
DOCA SDK documentation as the
next-reading link; an NGC catalog link alone is not documentation.
- For deployment routing, walk
TASKS.md ## recognize first only when the
container-versus-bare-metal path is not already established. If it
is established, proceed directly to the matching env workflow.
- Before a change recommendation on an installed/hardware target,
load the mandatory
Universal verification contract
and walk all applicable steps through an auditable green signal.
Hardware-touching answers also load the
hardware binding-layer command stanza.
The Stage-1 no-hardware path instead uses its own container
install/smoke green signal; never fabricate unavailable PCIe,
representor, or hardware evidence.
- For env workflows —
configure, test, debug, and
no-install — see TASKS.md. The build, modify,
and run anchors route to
doca-programming-guide/SKILL.md.
- Once the env is healthy, hand off to the selected deployment skill,
then
doca-programming-guide, then the matching library skill.
Both companion files cross-link to each other and to doca-public-knowledge-map whenever the right answer is "look it up in the public docs or the installed package layout" rather than "setup-specific guidance".
Related skills
doca-container-deployment/SKILL.md — the container deployment runtime (kubelet standalone + YAML pod-spec drop) for any DOCA service container on BlueField. ## recognize here routes to this skill when the developer's workload + system shape land on the container path.
doca-bare-metal-deployment/SKILL.md — the bare-metal hardware deployment runtime (host x86 OR BlueField Arm direct launch — systemd / tmux / direct invocation, hardware-resource binding, per-tenant isolation, restart discipline) for a DOCA-linked binary. ## recognize here routes to this skill when the developer's workload + system shape land on the bare-metal path.
doca-public-knowledge-map/SKILL.md — public DOCA documentation routing and the on-disk layout of an installed DOCA package. This skill defers all "where is X documented", "where on disk is Y", and "how do I check the installed version" questions to the knowledge-map.
doca-programming-guide/SKILL.md — general DOCA programming patterns once the env is healthy: the canonical pkg-config doca-<library> build pattern, the universal derive a custom first app from a sample workflow (with C / C++ + non-C tracks), the universal lifecycle, and the cross-library DOCA_ERROR_* taxonomy. Anything beyond "is the install healthy and the env ready" lives there.
doca-flow/SKILL.md — DOCA Flow on BlueField. Builds on this skill for env preparation and on doca-programming-guide for the universal first-app derivation, then layers Flow-specific overrides on top.
1---2license: Apache-2.0 AND CC-BY-4.03name: doca-setup4description: >5 Use this skill when the user is dealing with the DOCA environment6 around their workload — verifying an install is healthy, preparing7 the build env (pkg-config, headers, LD_LIBRARY_PATH, hugepages,8 devlink, representors), debugging env-class failures, deciding9 container-vs-bare-metal deployment shape, or reaching a DOCA install10 from a host that doesn't have one yet via the NGC DOCA container11 Stage-1 fallback. Trigger even when the user does not explicitly12 mention "DOCA setup" — typical implicit phrasings include "just13 got a BlueField, what now", "code is built, how to run it",14 "pkg-config can't find doca-flow", "no free 2048 kB hugepages",15 "representor X not found", "on a Mac and want to learn DOCA".16 Refuse and route elsewhere for library API specifics (Flow pipes,17 RDMA queues), the modify-a-sample first-app workflow or DOCA_ERROR_*18 program-side debugging, and "where is X documented" knowledge-map19 questions — those belong to other skills.20metadata:21 source: https://skillmd.com/skills/nvidia/doca-setup22 author: NVIDIA <doca-feedback@nvidia.com>23 kind: library24 doca_version: "3.5"25 tags:26 - doca27 - setup28 - environment29compatibility: >30 No DOCA install required to read this skill (it is an overlay loaded31 against any DOCA artifact skill); the validation steps within DO32 require a live DOCA install at /opt/mellanox/doca. The agent must33 have a target-host command channel or provide commands for the user34 to run and return their exact output; local shell access must not be35 assumed to reach the target.36---3738# DOCA setup3940**Where to start:** If the user's question is *deployment-shaped*41(*"how do I deploy"*, *"how do I run my DOCA workload"*, *"I just got a42BlueField, what now"*, *"my code is built, what next"*), walk43[`TASKS.md ## recognize`](references/recognize.md) **first**. It is the44bundle's front-door: it detects the system shape (host x86 / BlueField45Arm bare-metal / DPU-only / fresh laptop), asks the developer the46minimal set of questions needed to disambiguate, and routes to the47correct downstream skill — the container deployment path48(`doca-container-deployment/SKILL.md`),49the bare-metal hardware deployment path50(`doca-bare-metal-deployment/SKILL.md`),51or the no-hardware fallback52([`TASKS.md ## no-install`](TASKS.md#no-install)). The wrong failure53mode is to silently steer every developer onto containers because the54agent loaded that skill first; `## recognize` exists to prevent that.5556If the user does not have DOCA installed yet and the request is not57deployment-shaped, jump straight to58[`TASKS.md ## no-install`](TASKS.md#no-install) for the NGC container59path. Deployment-shaped requests still enter `## recognize` first,60which routes fresh-laptop cases to `## no-install`. Otherwise read61[`## When to load this skill`](#when-to-load-this-skill) to confirm the62question is env-class, then route to the section below that matches the63user's intent.6465**Do not use this skill for:**6667- Defining, validating or writing packet-steering rules (DOCA Flow pipes,68 matches, actions, entries), including when the user says the environment69 and versions are already verified. That belongs to `doca-flow`: name it70 and stop. Do not run `## recognize`, environment checks or the71 verification contract for it.72- Other library API work (RDMA queues, Comch channels, ...): the matching73 library skill.74- First-app derivation from a sample, build structure, or what a75 `DOCA_ERROR_*` means: `doca-programming-guide`.76- *"Where is X documented"*: `doca-public-knowledge-map`.7778## Examples7980The CLASSES this skill is built to handle, each with one worked example.81The skill must answer the *class*; the worked example is illustrative.8283## Instructions8485Follow [`## Loading order`](#loading-order). Open `TASKS.md` or86`CAPABILITIES.md` as named there. Do not open every companion file.8788- **"I want to deploy a DOCA workload — what is the right path?"** —89 worked example: *"I just got a BlueField; my code is built; what90 now?"* Resolved by the front-door decision tree in91 [`TASKS.md ## recognize`](references/recognize.md), which detects the92 system shape, asks the minimum residual question, and routes to93 either the container or the bare-metal deployment skill.94- **"Verify DOCA is installed and healthy."** — worked example:95 *"Is DOCA Flow actually available on this box?"* Resolved by the96 install-health snapshot in [`TASKS.md ## test`](TASKS.md#test) plus the97 version-detection rules in98 [`CAPABILITIES.md ## Capabilities and modes`](CAPABILITIES.md#capabilities-and-modes).99- **"I do not have DOCA installed — what now?"** — worked example: *"I'm100 on macOS and want to learn DOCA before I get a BlueField."* Resolved101 by [`TASKS.md ## no-install`](TASKS.md#no-install) (NGC DOCA container102 as universal Stage-1).103- **"Prepare the build environment for any DOCA library."** — worked104 example: *"`pkg-config --cflags doca-flow` returns nothing — what's105 missing?"* Resolved by the build-prep workflow in106 [`TASKS.md ## configure`](TASKS.md#configure) and the build-class107 error taxonomy in108 [`CAPABILITIES.md ## Error taxonomy`](CAPABILITIES.md#error-taxonomy).109- **"Prepare the runtime preconditions on a real DPU box."** — worked110 example: *"hugepages / representors / devlink — what's the minimum set111 before I run my first DOCA Flow program?"* Resolved by112 [`TASKS.md ## configure`](TASKS.md#configure) and the runtime113 observability rules in114 [`CAPABILITIES.md ## Observability`](CAPABILITIES.md#observability).115- **"Diagnose an env-class failure (install / build / runtime)."** —116 worked example: *"My DOCA Flow program built fine but says117 `pkg-config` cannot find it at runtime."* Resolved by the layered118 env-class debug workflow in [`TASKS.md ## debug`](TASKS.md#debug).119- **"Change something about the environment safely."** — worked120 example: *"I want to switch eswitch mode from legacy to switchdev."*121 Resolved by the safety constraints in122 [`CAPABILITIES.md ## Safety policy`](CAPABILITIES.md#safety-policy).123 Hugepage reservation, `mlxconfig` reset, and eswitch mode changes124 are host-global: they can disrupt live networking and representors,125 may require a reboot, and can be hard to undo. Obtain explicit user126 confirmation before recommending or executing them.127128If the question is library-API-shaped (Flow pipe construction, RDMA129queue setup, …) or program-shaped (how to build, modify a sample, debug130the program itself), route to131`doca-programming-guide/SKILL.md` or the132matching library skill instead — env-class only lives here.133134> **Route the *code* and stop — but keep the orientation.** For a135> library-API or program-shaped request, naming the owning skill and136> then answering it anyway is the failure mode to avoid: no137> pipe-construction steps, no queue-setup sketch, no example code, not138> even as a summary. This holds even when the user has already ruled139> the environment out ("versions are verified, I just need to define140> the rules") — that phrasing removes the env-class hook that would141> make it this skill's work.142>143> It is scoped to *that* work, and stops there. Saying where a144> component runs — which side of the BlueField / host split owns it,145> what the host talks to it over — is deployment orientation, which146> this skill owns as the front door and must still answer in full.147> The shared Core lifecycle is the same kind of orientation: name148> create → configure (properties, permissions, and tasks before149> start) → start → connect → use → stop → destroy, and that150> configuration MUST happen before `doca_ctx_start()`. Then hand the151> per-library calls to the owning skill. Handing off the API surface152> is not a reason to go quiet about the system shape around the153> object.154155## When to load this skill156157Load this skill when the user is dealing with the **environment around DOCA** — installing it, verifying the install is healthy, preparing the build / runtime preconditions, debugging env-class failures, figuring out *how to reach an install* from a host that doesn't have one yet, or asking a **deployment-shaped question** that hasn't yet been routed (containers vs. bare-metal) — the front-door routing decision lives here. Concretely:158159- The deployment-shape routing front door: *"I just got a BlueField, what now?"*, *"my code is built, how do I run it?"*, *"how do I deploy this?"* All of these load `## recognize` first so the agent does not silently push the user onto the wrong path.160- Verifying that the DOCA install is healthy and that the build environment can find it (`pkg-config`, headers, library paths).161- Preparing the runtime: hugepages, `devlink` device visibility, representor enumeration, kernel-module prerequisites.162- Diagnosing common setup-class failures: missing `*.pc` file, hugepages not mounted, representor not visible, header-vs-runtime version mismatch.163- The *I have no install yet* path: the user is on macOS, Windows, or a Linux box without DOCA, and needs to reach an environment where DOCA is actually installed. The canonical Stage-1 answer is the public NGC DOCA container on the catalog Tags page (macOS, Windows, or Linux with Docker Engine or Docker Desktop; no NVIDIA hardware required for the build / read / learn loop). See [`TASKS.md ## no-install`](TASKS.md#no-install).164165Do **not** load this skill for:166167- *"What is DOCA?", "where is the developer guide?", "where is the install layout documented?"* — those are routing questions; use `doca-public-knowledge-map/SKILL.md`.168- *"How do I derive a custom first application from a sample?", "how do I structure a DOCA build?", "what does `DOCA_ERROR_BAD_STATE` mean?"* — those are **programming-class** questions and live in `doca-programming-guide/SKILL.md`, which owns the universal `## modify` (first-app derivation), the canonical `## build` pattern, the universal lifecycle, and the cross-library `DOCA_ERROR_*` taxonomy.169- *Library-internal API questions* (Flow pipe construction, RDMA queue setup, etc.) — those belong in the matching library skill (e.g. `doca-flow/SKILL.md`). This skill stops at *"the install is healthy and the env is ready"*; it does not own program semantics.170171## What this skill provides172173This is a **thin loader**. The body keeps only the orientation needed to pick the right next file. The substantive env material lives in two companion files:174175- `CAPABILITIES.md` — what the install / build / runtime *environment* surface looks like: install profiles (`doca-all`, `doca-ofed`, `doca-networking`), where the build flavors (release vs trace) live on disk and how to point `LD_LIBRARY_PATH` at them, the env-side version-detection rules, the env-class error taxonomy (`pkg-config` not finding `doca-flow`, hugepages not reserved, representors not visible), what a healthy install looks like under observation, and the safety constraints on environment changes (hugepages global, `mlxconfig` reset, eswitch mode change).176- `TASKS.md` — env workflows: `recognize` (the front-door system-shape detect + dev-Q decision tree that routes deployment-shaped questions to either container or bare-metal), `configure` (env prep), `test` (install health snapshot), `debug` (env-class layered diagnosis), and `no-install` (the *I have no install yet* procedure with the NGC container as Path 0). Three other anchors (`build`, `modify`, `run`) route to `doca-programming-guide/SKILL.md`, which owns those verbs after the env / program split.177178This skill assumes nothing about whether DOCA is installed — the `## no-install` workflow exists precisely for the *fresh laptop* case.179180## Loading order1811821. Classify the question as programming,183 knowledge-map, beginner orientation, deployment routing, or184 environment work. Programming and knowledge-map questions route to185 their owning skill and stop here.1862. **For beginner orientation**, show the187 [`TASKS.md ## no-install` Stage 1 vs Stage 2 roadmap](TASKS.md#no-install)188 before any greeting, recommendation, or command. The opening must189 explicitly name **Stage 1 — container learning** and **Stage 2 —190 hardware runtime**; do not open by jumping directly to Docker or191 installation. Then route Stage 1/container learning to192 `## no-install` Path 0 and Stage 2/hardware runtime to193 `## no-install` Paths A/C, followed by `## recognize` before194 selecting container versus bare-metal. Ask one clarifying question195 when the stage is unknown; if it remains unknown or no reply is196 available in unattended execution, stop with197 `confirmation_required` and both paths explained rather than198 guessing. Include the authoritative199 [DOCA SDK documentation](https://docs.nvidia.com/doca/sdk/) as the200 next-reading link; an NGC catalog link alone is not documentation.2013. **For deployment routing**, walk202 [TASKS.md ## recognize](references/recognize.md) first only when the203 container-versus-bare-metal path is not already established. If it204 is established, proceed directly to the matching env workflow.2054. **Before a change recommendation on an installed/hardware target**,206 load the mandatory207 [Universal verification contract](CAPABILITIES.md#universal-verification-contract)208 and walk all applicable steps through an auditable green signal.209 Hardware-touching answers also load the210 [hardware binding-layer command stanza](references/hardware-binding.md).211 The Stage-1 no-hardware path instead uses its own container212 install/smoke green signal; never fabricate unavailable PCIe,213 representor, or hardware evidence.2145. **For env workflows — `configure`, `test`, `debug`, and215 `no-install` — see [TASKS.md](TASKS.md).** The `build`, `modify`,216 and `run` anchors route to217 `doca-programming-guide/SKILL.md`.2186. Once the env is healthy, hand off to the selected deployment skill,219 then `doca-programming-guide`, then the matching library skill.220221Both companion files cross-link to each other and to `doca-public-knowledge-map` whenever the right answer is *"look it up in the public docs or the installed package layout"* rather than *"setup-specific guidance"*.222223## Related skills224225- `doca-container-deployment/SKILL.md` — the container deployment runtime (kubelet standalone + YAML pod-spec drop) for any DOCA service container on BlueField. `## recognize` here routes to this skill when the developer's workload + system shape land on the container path.226- `doca-bare-metal-deployment/SKILL.md` — the bare-metal hardware deployment runtime (host x86 OR BlueField Arm direct launch — systemd / tmux / direct invocation, hardware-resource binding, per-tenant isolation, restart discipline) for a DOCA-linked binary. `## recognize` here routes to this skill when the developer's workload + system shape land on the bare-metal path.227- `doca-public-knowledge-map/SKILL.md` — public DOCA documentation routing and the on-disk layout of an installed DOCA package. This skill defers all *"where is X documented"*, *"where on disk is Y"*, and *"how do I check the installed version"* questions to the knowledge-map.228- `doca-programming-guide/SKILL.md` — general DOCA programming patterns once the env is healthy: the canonical `pkg-config doca-<library>` build pattern, the universal *derive a custom first app from a sample* workflow (with C / C++ + non-C tracks), the universal lifecycle, and the cross-library `DOCA_ERROR_*` taxonomy. Anything beyond *"is the install healthy and the env ready"* lives there.229- `doca-flow/SKILL.md` — DOCA Flow on BlueField. Builds on this skill for env preparation and on `doca-programming-guide` for the universal first-app derivation, then layers Flow-specific overrides on top.