Name: agent-platform-inference
Rating: 70.39999999999999 (1 reviews)
Author: google

agent-platform-inference

Connects to and performs inference with Google Cloud Agent Platform GenAI models, including First-Party Gemini models and Third-Party OpenMaaS models (Llama, DeepSeek, Qwen, etc.). Use when asked to perform inference, ask a model a question, run a test prompt, execute chat completions, or generate code for calling Gemini or OpenMaaS models, authenticate with GenAI SDK, OpenAI SDK, or legacy Agent Platform SDK, configure base URLs and global/regional endpoints, or troubleshoot 429 Resource Exhausted (DSQ), 400 User Validation, or 404 Not Found errors. Don't use for deploying models to endpoints or for running model evaluations.

Quality

87%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

75%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.

A highly actionable, well-structured skill that offloads code to real bundle scripts and includes explicit validation checkpoints for region/endpoint discovery. Its main weakness is conciseness — duplicated admonitions and repeated labels — plus a section-numbering inconsistency that slightly muddies navigation.

Suggestions

Remove the duplicated 'Alternative: Environment Variables' admonition (lines 384-388 repeat the same text twice) and consolidate the repeated 'Documentation:' labels.

Fix section numbering: the decision tree and §4 reference '[2. OpenMaaS Models]' and 'section 3', but the body has no §3 header and jumps from §2 to §4 — renumber or align the cross-references.

Consider moving the model-reference table and the detailed troubleshooting/error-code recipes into a references/ file to slim SKILL.md and deepen progressive disclosure.

Dimension	Reasoning	Score
Conciseness	Mostly efficient domain-specific guidance (DSQ, dedicated endpoint DNS, thinking-token pathology) but includes real padding: the 'Alternative: Environment Variables' admonition is duplicated verbatim (lines 384-388), 'Documentation:' labels repeat, and several admonition blocks could be tightened.	3 / 5
Actionability	Provides copy-paste-ready executable curl/Python/gcloud recipes plus complete offloaded scripts in scripts/ (verified present) and a verify_all.sh harness — fully executable guidance covering the common Gemini, OpenMaaS, and custom-endpoint cases.	5 / 5
Workflow Clarity	A clear Workflow Decision Tree sequences model-specification → SDK selection → troubleshooting, with validation checkpoints (region probe interpreted as 200/404/other, gcloud endpoint discovery before calling); capped below 5 by a section-numbering inconsistency (the tree and §4 reference 'section 3' / '[2. OpenMaaS Models]' but no §3 header exists — the doc jumps §2→§4).	4 / 5
Progressive Disclosure	Code is appropriately split into one-level-deep, clearly signaled scripts/ files (e.g. 'See [`scripts/gemini_genai_sdk.py`]...') that all exist; structure is good, though the model-reference table, troubleshooting, and full REST recipes are inlined in SKILL.md with no separate reference docs, a minor organization gap.	4 / 5
	Total	16 / 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.

A strong, specific description that names concrete actions, surfaces natural trigger terms including model families and error codes, and explicitly covers both what and when plus a negative boundary. It is clearly distinct from sibling deployment/eval skills.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions — 'performs inference', 'execute chat completions', 'generate code for calling Gemini or OpenMaaS models', 'authenticate with GenAI SDK, OpenAI SDK, or legacy Agent Platform SDK', 'configure base URLs and global/regional endpoints', 'troubleshoot 429... 400... or 404... errors' — giving comprehensive coverage.	5 / 5
Completeness	Explicitly answers what ('Connects to and performs inference with Google Cloud Agent Platform GenAI models...') and when ('Use when asked to perform inference, ask a model a question, run a test prompt...') with concrete trigger phrases, and adds a negative boundary ('Don't use for deploying models... or for running model evaluations').	5 / 5
Trigger Term Quality	Covers natural user phrases ('ask a model a question', 'run a test prompt', 'chat completions') plus model family names (Gemini, OpenMaaS, Llama, DeepSeek, Qwen), SDK names, and specific error codes (429, 400, 404) — comprehensive synonyms and technical triggers.	5 / 5
Distinctiveness Conflict Risk	Clear niche (Agent Platform GenAI inference) with distinct triggers (specific model families, SDKs, error codes) and an explicit exclusion ('Don't use for deploying models to endpoints') that separates it from the agent-platform-deploy skill — minimal conflict risk.	5 / 5
	Total	20 / 20 Passed

Validation

87%

Warnings & errors only

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

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (824 lines); consider splitting into references/ and linking	Warning
metadata_version	'metadata.version' is missing	Warning

	Total	14 / 16 Passed

Repository: google/skills
Path: skills/cloud/agent-platform-inference/SKILL.md
Commit: 6e209c4

Reviewed: 1 day ago

Table of Contents

Discovery Implementation Validation

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.