CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

doca-common

Use this skill whenever the user is doing hands-on DOCA programming on a BlueField DPU or ConnectX NIC and needs the foundation primitives every per-library context rests on — walking the doca_ctx lifecycle, discovering doca_dev / doca_devinfo and gating on doca_*_cap_* before trusting a feature, wiring doca_mmap / doca_buf_inventory / doca_buf for zero-copy I/O across libraries, driving doca_pe for completions, or DOCA Log's two-tier (--sdk-log-level vs app-side) model. Trigger even when the user does not say "DOCA Common" — typical implicit phrasings include "my tasks submit but nothing completes", "DOCA_ERROR_BAD_STATE from doca_ctx_start", "--sdk-log-level does nothing for my DOCA_LOG_DBG lines", "share a buf between doca_dma and doca_rdma", or "crashes far from the offending line". Refuse and route elsewhere for per-library questions in isolation (load doca-flow / doca-rdma / doca-eth alongside), installing DOCA (doca-setup), or doc lookup (doca-public-knowledge-map).

Quality

77%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./skills/doca-common/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

46%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body has the right shape for a foundation-routing skill — explicit scope, audience, load/not-load triggers, and a loading order — and includes a few concrete commands and paths. Its two real defects are repetition of the same routing message across five sections and, more seriously, that its entire disclosure strategy rests on two companion files that are neither named by path nor actually shipped in the bundle, making the core instructions unexecutable.

Suggestions

Name the companions with explicit resolvable paths (e.g. "see references/capabilities.md" / "references/tasks.md") and ship those files in the bundle — the Loading order, Instructions, and 'What this skill provides' sections all point at 'the capabilities companion' and 'the tasks companion' with no path, and no such files exist in the bundle.

Collapse the repeated routing into a single section: the capabilities-vs-tasks split, the doca-setup handoff, and the doc-lookup handoff are each restated in Instructions, When to load, What this skill provides, Loading order, and Related skills; merging them would cut roughly a third of the body.

Make each Loading order step executable on its own — include the exact `pkg-config --modversion doca-common` install-verification checkpoint inline and state what the agent should do if a companion file or an expected /opt/mellanox/doca path is missing.

DimensionReasoningScore

Conciseness

The body is honest that it is a "thin loader" and mostly avoids explaining general concepts, but the routing message is repeated across "Instructions", "When to load this skill", "What this skill provides", "Loading order", and "Related skills" — e.g. the two-tier log model, the doca-setup handoff, and the capabilities-vs-tasks split each appear three or more times, and the "Example questions this skill answers well" section runs ~45 lines of italicized prose that could be halved. It is mostly efficient domain signal but could be tightened by roughly a third; it does not reach anchor 2's 'several unnecessary explanations or padded sections' since no section is content-free.

3 / 5

Actionability

There are some genuinely concrete anchors — `pkg-config --modversion doca-common`, `/opt/mellanox/doca/samples/<library>/<sample>/*_main.c`, the full create→progress→stop verb chain, and named sibling-skill paths like `doca-setup/SKILL.md` — but the skill's primary instructed actions ("Open the tasks companion", "see the capabilities companion") name no file path, and no such files exist anywhere in the bundle (no references/, scripts/, or assets/ directories). The central instruction is therefore not executable as written, which is exactly the anchor-3 'missing key details' case rather than anchor 4's 'minor gaps'.

3 / 5

Workflow Clarity

"Loading order" gives a real 3-step sequence with an explicit question-type decision rule (what-does-Common-express → capabilities; do/configure/build/debug → tasks) and a precondition check (route to doca-setup if DOCA is not installed), so a coherent sequence is present. But the steps' destinations are unresolvable — the two companions have no paths and no files behind them — leaving the sequence present with key checkpoints implicit, which fits anchor 3 better than anchor 4's fully concrete per-step commands.

3 / 5

Progressive Disclosure

The thin-loader architecture is the right idea and the references are prominently signaled rather than buried, but scored against the actual bundle the disclosure structure is broken: the body's substantive material is promised to "two companion files" that are never given a path, and the bundle ships no references/, scripts/, or assets/ at all. An agent following the Loading order dead-ends immediately, which lands below the anchor-3 'references present but not clearly signaled' (they are signaled) and matches anchor 2's practical failure — the structure that clearly belongs in separate files is neither named nor present.

2 / 5

Total

11

/

20

Passed

Description

100%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

An exemplary description: concrete API-level capabilities, explicit 'Use when' triggers augmented with realistic implicit user phrasings and error strings, and explicit refuse-and-route boundaries against neighboring skills. The only nitpick is length (~170 words), but every clause is load-bearing trigger or routing signal rather than padding.

DimensionReasoningScore

Specificity

The description enumerates multiple concrete, API-level actions — "walking the doca_ctx lifecycle", "discovering doca_dev / doca_devinfo and gating on doca_*_cap_*", "wiring doca_mmap / doca_buf_inventory / doca_buf for zero-copy I/O", "driving doca_pe for completions", and the "two-tier (--sdk-log-level vs app-side)" log model — comprehensive coverage of the foundation surface with no vague filler. Anti-drift check: anchor 4 ('several specific actions; minor gaps') is too weak since every core primitive is named explicitly.

5 / 5

Completeness

Both halves are explicit: 'what' is the concrete foundation-primitives list, and 'when' opens with "Use this skill whenever the user is doing hands-on DOCA programming on a BlueField DPU or ConnectX NIC" and is reinforced with quoted implicit trigger phrases and a refuse-and-route clause. This matches the anchor-5 example structure (what + explicit when with concrete trigger phrases); it is not the anchor-4 case where 'when' is only generally stated.

5 / 5

Trigger Term Quality

It quotes the exact natural phrasings a stuck user would type — "my tasks submit but nothing completes", "DOCA_ERROR_BAD_STATE from doca_ctx_start", "--sdk-log-level does nothing for my DOCA_LOG_DBG lines", "share a buf between doca_dma and doca_rdma", "crashes far from the offending line" — plus device names (BlueField DPU, ConnectX NIC) and the pkg-config/module vocabulary. Anchor 4 ('a few natural terms missing') does not fit; coverage including implicit-synonym phrasings is comprehensive.

5 / 5

Distinctiveness Conflict Risk

Clear niche (cross-library foundation primitives under every doca_* context) with minimal conflict risk, actively defused by the explicit boundary "Refuse and route elsewhere for per-library questions in isolation (load doca-flow / doca-rdma / doca-eth alongside), installing DOCA (doca-setup), or doc lookup (doca-public-knowledge-map)". Anchor 4's 'minor overlap risk with closely related skills' is superseded because the overlapping skills are named and their traffic is explicitly rerouted.

5 / 5

Total

20

/

20

Passed

Validation

87%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 14 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

metadata_version

'metadata.version' is missing

Warning

metadata_field

'metadata' should map string keys to string values

Warning

Total

14

/

16

Passed

Repository

NVIDIA/skills

Path

skills/doca-common/SKILL.md

Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.