Agent skill · AI & Agents

doca-pcc-ztr-rttcc-algo

Use this skill when the user is doing hands-on deployment, tuning, or evaluation of the DOCA-shipped Zero-Touch RoCE RTT-based Congestion Control (ZTR RTTCC) reference algorithm on a BlueField-3 DPA — wiring `doca_pcc_dev_ztr_rttcc_algo` into the shipped DOCA PCC sample, picking a variant (vanilla / PM / RX-rate / multipath / window-probeless) at DPACC build time, tuning host-set parameters, or diagnosing `DOCA_PCC_DEV_STATUS_FAIL` from the algorithm. Trigger even when the user does not say 'DOCA PCC' or 'ZTR RTTCC' — typical implicit phrasings: 'my RoCE-v2 flows aren't being throttled', 'PC

NVIDIAgithub.com/NVIDIAGitHub ↗
claude-codecodexApache-2.0
Install
npx skills add NVIDIA/skills --skill doca-pcc-ztr-rttcc-algo --agent claude-code

Same command for any agent — swap --agent for codex, cursor, copilot.

Facts
Files in the skill folder: 8
SKILL.md size: 19 KB
Bundled scripts: none
Requires: > Requires DOCA SDK installed at /opt/mellanox/doca on Linux (Ubuntu 22.04/24.04 or RHEL/SLES) with a BlueField-3 DPU…
Path: skills/doca-pcc-ztr-rttcc-algo/SKILL.md
Open the folder on GitHub →
Where it comes from
Stars: 2,789
Language: Python
Read our review of the source →

Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.

Review
written from the skill's own SKILL.md · Aug 5, 2026

What it does

The skill serves as a thin loader that helps the agent guide users through deploying and configuring the DOCA-shipped ZTR RTTCC algorithm on a BlueField-3 DPU. It directs users to pick a variant (vanilla / PM / RX-rate / multipath / window-probeless) at DPACC build time, wire the algorithm into the DOCA PCC sample, and tune host-set parameters or diagnose DOCA_PCC_DEV_STATUS_FAIL from the algorithm. It references workflows for install, modify, run, test, and use, and points to CAPABILITIES.md and TASKS.md for specifics. The agent should route users to the installed library and header, and prescribe in-place edits to the shipped PCC application as documented in the README, rather than implementing the algorithm itself.

How it works

  • The skill acts as guidance to deploy and evaluate the shipped ZTR RTTCC algorithm on a BlueField-3 DPU carrying RoCE-v2 traffic.
  • It covers selecting a variant at DPACC build time and wiring the user algorithm into the DOCA PCC sample by patching the user-init / user-set-algo-params callbacks per the shipped README.
  • It instructs tuning host-side parameters via doca_pcc_dev_set_ztr_rttcc_params and examining the parameter surface described in the provided CAPABILITIES and TASKS.
  • It emphasizes observing whether the algorithm modulates traffic, using the doca-pcc-counters tool for read-only inspection, and deciding if a custom algorithm should be used.
  • It does not ship or author the algorithm body; it routes to the installed library and header (doca_pcc_dev_ztr_rttcc_algo.*) and to the host-side doca-pcc framework.

When to use it

Load this skill during hands-on deployment, tuning, or evaluation of the DOCA-shipped ZTR RTTCC reference algorithm on a BlueField port with RoCE-v2 traffic. Use when deciding baseline vs custom algorithm, wiring into the PCC sample, selecting a variant, tuning host parameters, observing modulating behavior, or determining replacement with a custom algorithm.

What it can touch

  • The host-side PCC framework (doca-pcc) and its integration steps as described in TASKS.md and CAPABILITIES.md.
  • The DOCA library and header related to the ZTR RTTCC algorithm: the public DPA-side API surface such as doca_pcc_dev_ztr_rttcc_init, doca_pcc_dev_ztr_rttcc_algo, doca_pcc_dev_set_ztr_rttcc_params, and related getters/counters as listed in CAPABILITIES.md.
  • The variant surface (vanilla / path-migration / RX-rate / multipath / window-probeless) and the single public symbol doca_pcc_dev_ztr_rttcc_algo that is selected at compile time on the DPA side.

Caveats

  • Requires DOCA SDK installed at /opt/mellanox/doca with a BlueField-3 DPU exposing the DPA processor, firmware custom-PCC slot enabled, a matched DPACC compiler, and live RoCE-v2 traffic.
  • The skill is guidance-focused and does not ship the algorithm body or a standalone PCC application; it prescribes edits to the existing installed layout and workflow steps per README.
  • It explicitly routes questions about writing a custom algorithm from scratch or read-only PCC counter inspection to other skills or documentation.
From the SKILL.md

# DOCA PCC ZTR RTTCC Algorithm **Where to start:** This skill assumes DOCA is already installed, the user's BlueField has a DPA processor that the host can see through DOCA (a BlueField-3-generation device per the README), the BlueField firmware has the custom-PCC slot enabled, the DPACC compiler is installed at a matched version per the DOCA Compatibility Policy, and the user is doing **hands-on

More from skills
All skills →
About this skill
What does the doca-pcc-ztr-rttcc-algo skill do?

Use this skill when the user is doing hands-on deployment, tuning, or evaluation of the DOCA-shipped Zero-Touch RoCE RTT-based Congestion Control (ZTR RTTCC) reference algorithm on a BlueField-3 DPA — wiring `doca_pcc_dev_ztr_rttcc_algo` into the shipped DOCA PCC sample, picking a variant (vanilla / PM / RX-rate / multipath / window-probeless) at DPACC build time, tuning host-set parameters, or diagnosing `DOCA_PCC_DEV_STATUS_FAIL` from the algorithm. Trigger even when the user does not say 'DOCA PCC' or 'ZTR RTTCC' — typical implicit phrasings: 'my RoCE-v2 flows aren't being throttled', 'PC

How do I install it?

Run `npx skills add NVIDIA/skills --skill doca-pcc-ztr-rttcc-algo --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.

Keep going