Doca Comm Channel Admin
NVIDIAOfficialCreated by NVIDIA
- Skill
- Tool integration
- Software development
- Apache-2.0
- Medium risk
AutoGineer summary
Written by AutoGineer from the skill’s own files, not by the author
This skill provides guidance for using the doca_comm_channel_admin binary to perform a read-only, single-shot inventory of DOCA Comm Channel (comch) servers and connections. It is designed for operators and agents needing to verify channel health or debug connectivity issues between a host and a DPU. The tool scans all comch-capable devices and outputs status tables, serving as an observability primitive to complement program-side debugging.
What it does
- Invokes the doca_comm_channel_admin binary for a single-shot scan.
- Enumerates comch-capable devices on the current system.
- Decodes and prints server and connection tables including PIDs and PCIe addresses.
- Provides a read-only view of channel state for diagnostic purposes.
- Facilitates cross-checking between program-reported status and actual hardware state.
When to use it
- When you need to list currently active comch channels on a host or DPU.
- When debugging a silent stall or connection error in a Comch-using application.
- When verifying that a channel is healthy before declaring a service ready.
- When you need to capture a snapshot of channel state as evidence for deeper debugging.
Advantages
- Provides an external, objective view of channel state independent of the application.
- Simple, zero-argument execution model for quick inventory.
- Helps distinguish between application-level logic errors and underlying channel connectivity issues.
- Standardized approach for observability across different DOCA environments.
Usage examples
- “Which comch servers and connections are currently visible?”
- The agent executes the doca_comm_channel_admin tool to output the current SERVERS and CONNECTIONS tables.
- “My program reports CONNECTED but the admin tool lists zero channels. What is happening?”
- The agent uses the tool to verify the hardware-level view, then investigates representor binding or potential mismatches between the program and the admin tool's perspective.
Install
Download the skill as a ZIP:
https://mp-staging.autogineer.com/agent-skills/nvidia/doca-comm-channel-admin/downloadUnzip it into the cross-client skills folder of your project:
mkdir -p .agents/skills && unzip doca-comm-channel-admin.zip -d .agents/skills
Documentation by NVIDIA
DOCA Comm Channel Admin Tool
Examples
"Which comch servers and connections are currently visible?" — this is in scope. Stay on the admin CLI; do not open a new comch channel from this skill.
"The program reports CONNECTED but the admin tool lists zero channels." — cross-check representor binding before rewriting the comch app.
Instructions
Follow Loading order below. Open the tasks companion for run / debug. Open the capabilities companion for table columns and error layers. Do not open every companion file.
Actual binary contract.
doca_comm_channel_adminis one zero-application-argument, read-only scan-and-print operation. It scans every comch-capabledoca_devon the current side viaresourcedump(MFT) and prints SERVERS and CONNECTIONS tables. It has no list, inspect, device-scope, drain, restart, or other application operation. This skill does not retain conceptual workflows under invented command names.
Where to start: This is a tool skill for invoking the DOCA Comm
Channel Admin Tool — the read-only inventory CLI counterpart
to the doca-comch/SKILL.md library. The
shipped doca_comm_channel_admin binary takes no arguments
beyond ARGP defaults (--help, --version, --log-level,
--sdk-log-level, --json) and performs one inventory pass
per invocation: it walks every doca_dev on this side, filters
to comch-capable devices, shells out to resourcedump (MFT) on
each, and prints two ASCII tables (SERVERS and CONNECTIONS).
There is no list subcommand, no inspect subcommand, no
drain flag, and no restart flag — those are not part of the
tool's surface. Open the tasks companion and start at
## run for the single-shot invocation, or
## debug when the user reports the tool sees
a different channel set than the program. Open
the capabilities companion when the question is what
the printed tables actually mean and what is not in this
tool's scope. If the user has not installed DOCA yet, route to
doca-setup/SKILL.md first; if the user
needs MFT (resourcedump on PATH with the privilege documented
for the installed release), doca-setup + doca-public-knowledge-map
cover that. If the user is holding pre-2.5 docs that mention
"Comm Channel", route to
doca-comch/SKILL.md
for the rename rule. If the user wants to change channel
state, route to the program-side
reconnect lifecycle in doca-comch/SKILL.md
or to BlueField mode / driver reload in
doca-setup/SKILL.md +
doca-hardware-safety/SKILL.md — not
to this tool.
Example questions this skill answers well
The CLASSES of admin-tool 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.
"Which comch servers and connections are currently visible?" — worked example: "print every server and connection row visible on this side". Answered by the scan-and-print surface in the capabilities companion
the single invocation in the tasks companion.
"What does the tool report for this server or connection?" — locate the matching row in the SERVERS or CONNECTIONS table; there is no second per-channel query. Answered by the capabilities companion
the tasks companion.
"How do I know the admin tool's view matches what my Comch program sees?" — worked example: "the program reports CONNECTED but the admin tool lists zero channels". Answered by the cross-checking pattern in the tasks companion and the representor-binding layer in the capabilities companion.
"Is this admin tool on my installed DOCA version, and does it match the comch library version?" — worked example: "is the tool available on DOCA 2.4". Answered by the overlay in the capabilities companion, which redirects to the canonical
doca-version/SKILL.mdrules and adds the Comm Channel Admin Tool specifics."The tool prints nothing — is the install broken or is there genuinely no channel?" — worked example: "
listreturned an empty result on a host with a known-good Comch client". Answered by the empty-output interpretation rules in the tasks companion + the capabilities companion.
Audience
This skill serves external operators and AI agents who need to
inventory a host-DPU comch channel from the outside
— after the doca-comch/SKILL.md library
has been used to create the channel from a program. Concretely:
A platform operator who runs a Comch-using service on BlueField and needs to confirm the channel is healthy before declaring the service ready.
A developer of a Comch consumer who sees
DOCA_ERROR_AGAINor a silent stall on the program side and wants to read the channel's state from outside the program rather than guessing.An AI agent capturing the external server/connection rows before recommending a program-side code or lifecycle change.
It is not for users debugging the admin tool itself, not a
substitute for the live public DOCA Comm Channel Admin Tool guide,
and not the right place for users learning the comch API —
that audience belongs in
doca-comch/SKILL.md.
The tool is shipped as a CLI binary under
/opt/mellanox/doca/tools/, not a library you link against. The
skill uses the same kind: tool three-file shape as the rest of
the bundle so the agent's task-verb contract
(configure / build / modify / run / test / debug) is uniform
across libraries, services, and tools.
When to load this skill
Load this skill when the user is — or the agent needs to — invoke the DOCA Comm Channel Admin Tool on a real host or BlueField Arm with DOCA installed (or inside the public NGC DOCA container with the right device passthrough). Concretely:
Listing currently active comch channels on a host or DPU.
Reading the row for one named server or connection from the complete zero-argument scan.
Cross-checking the admin tool's view against the program-side connection callback state when the two appear to disagree.
Capturing a side-effect-free channel snapshot as prerequisite evidence for a later debug session that crosses program / channel / driver layers.
Do not load this skill for general DOCA orientation, the comch
programming API, library install, or comch protocol design. For
those, route to
doca-public-knowledge-map/SKILL.md,
doca-comch/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 the Comm Channel Admin Tool reports: the two read-only tables, the version-availability overlay that redirects to
doca-version/SKILL.md, the layered error taxonomy (tool-not-installed / device-binding / channel-discovery / channel-state-stuck / permission / version / cross-cutting), the tool's role as an observability primitive fordoca-comch/SKILL.mddebug sessions, and the read-only safety policy.the tasks companion — step-by-step workflows for the in-scope task verbs:
configure(route to install),build(route to install),modify(refuse),run(one scan-and-print),test(cross-check the printed rows),debug(the layered diagnosis ladder), plus aDeferred task verbsblock and aCommand appendixthat honors the bundle'sdoca-structured-tools-contract/SKILL.mdpreamble.
The skill assumes a host or BlueField where DOCA is already installed (or the public NGC DOCA container is running with the right device passthrough) and the operator has whatever privileges the public DOCA Comm Channel Admin Tool guide requires.
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, or output column names. The public DOCA Comm Channel Admin Tool guide on
docs.nvidia.comand the installed--helpon the user's version are the joint source of truth; copying them here pins the skill to one release and silently rots when the tool evolves. The skill routes the agent at those sources instead.Pre-baked example output. Output is install-, version-, and channel-state-specific. A captured example will mislead an operator on a different platform / state.
Wrappers, parsers, or scripts in any language that consume the admin tool's output. The output format is documented; users who want to script against it should read the live guide and write the parser against their installed version.
A
samples/orreference/subtree. This is a thin loader for a documented CLI; substantive material lives on the public page and in--help.
Loading order
Read this skill first to confirm the user's question is in scope (the user wants to inventory a comch channel from the outside, not learn the comch API).
For what the tool reports, version availability, the layered error surface, observability, and safety posture, see the capabilities companion.
For the single invocation and cross-check workflow —
configure,build,modify,run,test,debug, plus theCommand appendix— see the tasks companion.
Related skills
doca-comch/SKILL.md— the library whose channels this tool inventories. Pair them in every triage session: the program-side connection callback and the admin tool's channel state are the two halves of the same picture.doca-public-knowledge-map/SKILL.md— routing to the public DOCA Comm Channel Admin Tool guide and the rest of the public DOCA documentation set.doca-version/SKILL.md— canonical DOCA version-handling rules. The## Version compatibilitysection in the capabilities companion is a concise overlay that redirects here for the body.doca-structured-tools-contract/SKILL.md— the bundle's detect → prefer → fall back → report contract for structured helper tools. The Command appendix in the tasks companion honors this contract.doca-setup/SKILL.md— env preparation, install verification, representor visibility checks, and the I have no install yet path with the public NGC DOCA container. This skill assumes its preconditions are satisfied.doca-debug/SKILL.md— the cross-cutting debug ladder. The Comm Channel Admin Tool slots in at the runtime layer as the read-only inventory surface before any code change is recommended.
Files
- ATTRIBUTION.mdAdded by AutoGineer927 B
- BENCHMARK.md10.5 KiB
- CAPABILITIES.md12.6 KiB
- LICENSELicence file added by AutoGineer11.2 KiB
- SKILL.md11.9 KiB
- SKILLCARD.yaml2.6 KiB
- TASKS.md15.0 KiB
- evals/evals.json2.5 KiB
- skill-card.md4.7 KiB
- skill.oms.sig5.1 KiB
BENCHMARK.md
Skill Benchmark: doca-comm-channel-admin
✅ Overall verdict: PASS — Recommended for publication
Publication Recommendation
Recommended for publication based on the completed evaluation evidence in this report.
Evaluation Metadata
Skill:
doca-comm-channel-adminValidation status:
passedEvaluation date: 2026-10-28
Evaluator version:
1.6.2Agents: Claude Code (
aws/anthropic/bedrock-claude-opus-5-5), Codex (openai/openai/gpt-6.1-sol)Tasks: 4 evaluation tasks (3 positive, 1 negative)
Dataset digest:
sha256:9d2e8574f2afb5f9a860008df26c43b3364e7a22f743c6823e50820b6cd7785a(skill-evaluator-dataset-snapshot/1)Attempts per task: 1
Environment:
k8s-sandboxTier 2 evidence: required for publication
Tier 3 evidence: required for publication
Each task attempt ran in its own isolated sandbox pod.
Tier 1 LLM Provenance
SkillSpector provider, model, analyzer revision, and determinism evidence retained for this scan.
| Skill | Provider | Adapter | Analyzer | Model | Revision | Determinism |
|---|---|---|---|---|---|---|
| doca-comm-channel-admin | nv_inference | openai | mcp_tool_poisoning | aws/anthropic/bedrock-claude-opus-5-5 | 2.13.1 | nondeterministic / provider_defaults |
| doca-comm-channel-admin | nv_inference | openai | semantic_developer_intent | aws/anthropic/bedrock-claude-opus-5-5 | 2.13.1 | nondeterministic / provider_defaults |
| doca-comm-channel-admin | nv_inference | openai | semantic_quality_policy | aws/anthropic/bedrock-claude-opus-5-5 | 2.13.1 | nondeterministic / provider_defaults |
| doca-comm-channel-admin | nv_inference | openai | semantic_security_discovery | aws/anthropic/bedrock-claude-opus-5-5 | 2.13.1 | nondeterministic / provider_defaults |
| doca-comm-channel-admin | nv_inference | openai | meta_analyzer | aws/anthropic/bedrock-claude-opus-5-5 | 2.13.1 | nondeterministic / provider_defaults |
Controls for
doca-comm-channel-admin: temperature=not set (source=provider_default, forwarded=no); seed=not set (source=unset, forwarded=no); reasoning_effort=not set (source=provider_default, forwarded=no)
What This Report Answers
The three-tier evaluation checks whether the skill:
is safe to use;
produces correct answers;
is discovered and activated when needed;
helps the agent complete the user's goal and expected workflow; and
avoids wasted skill and tool usage.
Results at a Glance
| Measure | Claude Code (Baseline → Skill Uplift) | Codex (Baseline → Skill Uplift) |
|---|---|---|
| Overall | 92.2% — baseline ran, but no comparable score was available; uplift unavailable | 93.5% — baseline ran, but no comparable score was available; uplift unavailable |
| Security | 100.0% → 100.0% (±0.0 points) | 100.0% → 100.0% (±0.0 points) |
| Correctness | 25.0% → 95.0% (+70.0 points) | 55.0% → 100.0% (+45.0 points) |
| Discoverability | 100.0% — baseline ran, but no comparable score was available; uplift unavailable | 95.0% — baseline ran, but no comparable score was available; uplift unavailable |
| Effectiveness | 12.5% → 75.0% (+62.5 points) | 37.5% → 93.8% (+56.3 points) |
| Efficiency | 91.0% — baseline ran, but no comparable score was available; uplift unavailable | 78.8% — baseline ran, but no comparable score was available; uplift unavailable |
How to read this table: baseline is the same task attempted without the target skill. Scores are rounded to one decimal; threshold-adjacent values use additional precision so their displayed band matches the verdict. Uplift is derived from those displayed scores and shown in percentage points.
Example: 47.0% → 92.0% (+45.0 points) means the skill-assisted run scored 92.0%, 45.0 percentage points above its 47.0% no-skill baseline.
A partial dimension was calculated from only the available configured signals; review the detailed report before relying on it.
Comparison details
Shared-source comparisons use the same signals and weights on both conditions. They do not replace the full dimension scores above or change the publication verdict.
| Agent | Dimension | Basis | Baseline → With skill (uplift) | Explanation |
|---|---|---|---|---|
| claude-code | Discoverability | Dimension comparison | N/A | Discoverability is not applicable to the no-skill baseline; there is no installed skill to discover. |
| claude-code | Efficiency | Token-only comparison | 98.5% → 95.2% (-3.3 points) | Only Token Efficiency is comparable; the full dimension scores use different evaluator sources. |
| codex | Discoverability | Dimension comparison | N/A | Discoverability is not applicable to the no-skill baseline; there is no installed skill to discover. |
| codex | Efficiency | Token-only comparison | 89.3% → 88.2% (-1.1 points) | Only Token Efficiency is comparable; the full dimension scores use different evaluator sources. |
Token Usage
Actual Tier 3 execution usage is reported for every observed agent/case pair and both conditions.
| Agent | Dataset case | With skill | Without skill | Delta | Change | Coverage |
|---|---|---|---|---|---|---|
| claude-code | All cases | 730,124 | 321,949 | +408,175 | +126.78% | skill 4/4; base 4/4 |
| claude-code | doca-comm-channel-admin.explicit-positive.v1 | 208,454 | 107,463 | +100,991 | +93.98% | skill 1/1; base 1/1 |
| claude-code | doca-comm-channel-admin.implicit-positive.v1 | 242,827 | 71,736 | +171,091 | +238.50% | skill 1/1; base 1/1 |
| claude-code | doca-comm-channel-admin.implicit-positive.v2 | 243,021 | 71,739 | +171,282 | +238.76% | skill 1/1; base 1/1 |
| claude-code | doca-comm-channel-admin.negative.v1 | 35,822 | 71,011 | -35,189 | -49.55% | skill 1/1; base 1/1 |
| codex | All cases | 299,859 | 123,137 | +176,722 | +143.52% | skill 4/4; base 4/4 |
| codex | doca-comm-channel-admin.explicit-positive.v1 | 97,146 | 30,516 | +66,630 | +218.34% | skill 1/1; base 1/1 |
| codex | doca-comm-channel-admin.implicit-positive.v1 | 71,266 | 26,000 | +45,266 | +174.10% | skill 1/1; base 1/1 |
| codex | doca-comm-channel-admin.implicit-positive.v2 | 96,840 | 21,438 | +75,402 | +351.72% | skill 1/1; base 1/1 |
| codex | doca-comm-channel-admin.negative.v1 | 34,607 | 45,183 | -10,576 | -23.41% | skill 1/1; base 1/1 |
| ALL AGENTS | Dataset aggregate | 1,029,983 | 445,086 | +584,897 | +131.41% | skill 8/8; base 8/8 |
Prompt tokens include cached reads, so total tokens are prompt + completion (cached is not added twice). The Efficiency score uses (prompt - cached) + completion. Change is 100 * (with - without) / without; negative means fewer tokens. N/A comparisons include their reason in the coverage column; coverage is never estimated.
Tier Status
| Tier | Purpose | Status | Evidence |
|---|---|---|---|
| Tier 1 | Static validation | PASSED WITH OBSERVATIONS | 11 validator(s); 12 finding(s) |
| Tier 2 | Semantic deduplication | PASSED WITH OBSERVATIONS | 2 validator(s); 3 finding(s) |
| Tier 3 | Live agent evaluation | PASS | 2 agent(s); 4 task(s) |
Findings and Observations
<details> <summary>Show detailed findings and successful checks</summary>
CRITICAL CONTENT_DEDUP/llm_error: LLM analysis failed for a content cluster (
skills/tools/doca-comm-channel-admin)HIGH DUPLICATE/duplicate: Duplicate content found within SKILL.md: "## Examples" in SKILL.md (lines 3-11) vs "## Example questions this skill answers well" in SKILL.md (lines 58-95) (
SKILL.md:3)HIGH DUPLICATE/duplicate: Duplicate content found across CAPABILITIES.md and SKILL.md and TASKS.md: "# DOCA Comm Channel Admin Tool — Capabilities" in CAPABILITIES.md (lines 1-20) vs "## Capabilities and modes" in CAPABILITIES.md (lines 45-87) vs "## Instructions" in SKILL.md (lines 12-57) vs "# DOCA Comm Channel Admin Tool — Tasks" in TASKS.md (lines 1-31) vs "## run" in TASKS.md (lines 114-147) (
CAPABILITIES.md:1)MEDIUM QUALITY/quality_discoverability: Description contains vague words (
skills/tools/doca-comm-channel-admin/SKILL.md)MEDIUM SCHEMA/metadata_key_style: Metadata key 'doca_version' is not kebab-case (
skills/tools/doca-comm-channel-admin/SKILL.md)10 additional finding(s) are available in the full evaluation artifacts.
</details>
Scoring Methodology
<details> <summary>Show dimension definitions, source signals, and thresholds</summary>
| Dimension | Question | Scored signals |
|---|---|---|
| Security | Is it safe to use? | security (100%) |
| Correctness | Is the answer correct? | accuracy (100%) |
| Discoverability | Was the right skill loaded when needed? | skill_execution (100%) |
| Effectiveness | Did the skill help complete the task? | goal_accuracy (50%) + behavior_check (50%) |
| Efficiency | Did it avoid wasted tool calls and token usage? | skill_efficiency (50%) + token_efficiency (50%) |
Dimension bands: PASS at 50% or above; NEUTRAL from 40% to below 50%; FAIL below 40%.
Overall Tier 3 lift: PASS at +5 points or more; FAIL at -10 points or less; values between those bands are NEUTRAL.
Overall verdict: PASS only when every configured dimension passes for at least one supported agent. Lift is reported as diagnostic evidence and does not override this gate.
The 50% attempt pass threshold is a separate per-task gate; it is not the dimension pass threshold.
Effectiveness is the equal-weight mean of goal completion (
goal_accuracy) and expected workflow adherence (behavior_check).Efficiency is 50% tool-call productivity (the backward-compatible
skill_efficiencywire id) and 50%token_efficiency. Positive-case skill routing is scored under Discoverability, not Efficiency; a negative case without a routing target is N/A. N/A sources are omitted, remaining weights are renormalized, and the dimension is marked partial.
Signals present in this run:
security(Security): unsafe operations, secret leakage, and unauthorized access.skill_execution(Skill Execution): whether the expected skill was selected, decoys were avoided, and the workflow executed.skill_efficiency(Tool Productivity): tool-call productivity (legacy wire id; routing is scored under Discoverability).accuracy(Accuracy): final-answer correctness against the reference answer.goal_accuracy(Goal Accuracy): whether the user's goal was achieved.behavior_check(Behavior Check): whether the expected workflow behavior was followed.token_efficiency(Token Efficiency): actual uncached prompt plus completion usage (50% of Efficiency).
</details>
Freshness
Regenerate this benchmark when the skill, evaluation dataset, target agent/model, evaluator version, environment, or scoring policy changes.
Requirements and compatibility
Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Goose, Amp, Roo Code, JetBrains Junie, Factory Droid, Devin (ex-Windsurf), Mistral Vibe, OpenHands, Letta
“Works with” means the tool reads the open Agent Skills format; a skill written for one tool may use features others lack.
The author says: Requires DOCA SDK installed at /opt/mellanox/doca on Linux (Ubuntu 22.04/24.04 or RHEL/SLES) with the Comm Channel tooling subpackage so the CLI is present under /opt/mellanox/doca/tools/. Runs from either the x86/Arm host or the BlueField Arm side; a live host↔DPU comch channel created via the doca-comch library is needed for non-empty tables. The validation steps require a live DOCA install.
Security
What our scanner noticed· Medium risk
- Large base64-encoded blobs
A static scan is not a guarantee. Skills are third-party content: review a skill before you let an agent use it.
Attribution and licence
- Created by
- NVIDIA
- Publisher
- NVIDIA (Official)
- Source
- NVIDIA/skills · skills/doca-comm-channel-admin · commit fce3cca
- Licence
- Apache License 2.0 (redistribution allowed)
- LICENSE (added by AutoGineer from LICENSE-APACHE)
- ATTRIBUTION.md (written by AutoGineer)
- Synced
AutoGineer is not affiliated with or endorsed by NVIDIA.
Download ZIP
The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download doca-comm-channel-admin.zip