CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

microsoft-foundry

Build, deploy, evaluate, optimize, fine-tune, and manage Microsoft Foundry agents, models, and resources end to end. USE FOR: foundry, azd ai agent, azd provision/deploy, hosted agent scaffold/develop/run/deploy/troubleshoot, prompt agent, managed harness agent, Copilot harness, voice agent, create/update/invoke/evaluate agent, add tool to agent, agent insights, batch/continuous eval, continuous monitoring, agent CI/CD, optimize/improve prompt, optimize agent instructions, Agent Optimizer, dataset curation from traces, deploy model, model fine-tuning (SFT/DPO/RFT), Foundry project, RBAC, role assignment, permissions, quota, capacity, region, deployment failure, AI Services, create Foundry resource, knowledge index, customize deployment, onboard, availability, training-data, grader, distillation, large file upload. DO NOT USE FOR: Azure Functions, App Service, general Azure deploy (use azure-deploy), general Azure prep (use azure-prepare).

Quality

90%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Failed to scan

The risk profile of this skill

SKILL.md
Quality
Evals
Security

Quality

Content

77%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 dense, highly actionable orchestrator document with excellent sequencing, validation checkpoints, and error-recovery loops, and virtually no filler. Its weaknesses are token efficiency (systematic repetition of prerequisite prefixes and an over-long inline context-resolution section) and progressive disclosure (25 of 28 linked sub-skill documents are absent from the bundle, leaving most routing links unresolved).

Suggestions

State the "dependency check and setup → azd-guidance" prerequisite once above the lifecycle table and drop it from every row — the ~12 repeated link chains are the single largest avoidable token cost in the body.

Move the seven-step Common Project Context Resolution into a reference file (e.g. references/agent-context-resolution.md) and keep a short summary plus link in SKILL.md, mirroring the existing agent-metadata-contract.md pattern.

Ensure the referenced sub-skill files ship with the skill bundle — 25 of 28 .md links (deploy, invoke, observe, finetuning, rbac, etc.) currently point to files that do not exist, breaking the router's navigation; also consider flattening the two-level-deep voice-agent/subtypes.md reference.

DimensionReasoningScore

Conciseness

The body is almost entirely Foundry-specific operational knowledge Claude would not already know (no concept-explanation padding), but it could be meaningfully tightened: the ~12 rows of the Agent Development Lifecycle table each repeat the identical "[dependency check and setup](#dependency-check-and-setup) → [azd-guidance](...) →" prefix, the MANDATORY prerequisite rules are restated across three Pre-Execution subsections, and the ~100-line seven-step context resolution runs long. This fits level 3's "mostly efficient but ... could be tightened" rather than level 4's "minor instances" — the repetition is systematic, not incidental — while level 2's "several unnecessary explanations" overstates it since nothing explains known concepts.

3 / 5

Actionability

Concrete, executable guidance throughout: copy-paste-ready commands ("./scripts/check-and-setup-dependencies.sh", the full "az cognitiveservices account show --query ..." diagnostic), exact environment-variable tables (AZURE_AI_PROJECT_ENDPOINT, AGENT_<SERVICE>_NAME), exact tool names (ask_user, evaluator_catalog_get, prompt_optimize), and precise ordered decision rules (metadata file selection order, agent-type classification order). This matches level 5's "fully executable ... specific examples cover the common cases" — for an orchestrator/instruction skill there are no pseudocode substitutes, satisfying the code_vs_instruction scoring note.

5 / 5

Workflow Clarity

Multi-step processes are explicitly sequenced with validation checkpoints and feedback loops: Steps 1–7 each define fallback ordering with "verify with evaluator_catalog_get before treating them as remote evaluators", "Persist synced ... references ... only after remote lookup or registration succeeds", and "If azd and metadata both provide the same value and they differ, stop and ask"; the Network Isolation Errors section is a full diagnose → confirm with management-plane call → interpret branches → remediate loop; and metadata writes (a batch/overwrite-risky operation) carry guardrails ("ask before refreshing or overwriting them", "Never copy or merge environments across sibling metadata files"). This matches level 5; level 4's "minor validation gaps" understates the pervasive checkpointing.

5 / 5

Progressive Disclosure

The router design is well-signaled (a When-to-Use table plus lifecycle tables with markdown links), but scoring against the actual bundle reveals 25 of the 28 referenced .md files are missing (e.g. foundry-agent/deploy/deploy.md, models/deploy-model/SKILL.md, quota/quota.md) — only the 3 references/ files exist — so most navigation dead-ends. Additionally, one referenced path sits two levels deep (foundry-agent/create/references/voice-agent/subtypes.md is linked directly from SKILL.md) and ~100 lines of context-resolution detail (Steps 1–7) are inlined in SKILL.md where the skill already uses reference files for comparable material. This lands at level 3's "some structure but could be better organized" — above level 2 ("minimal structure ... references are buried" does not hold; structure and signaling are good) but below level 4 ("minor organization gaps" — broken reference targets are not minor).

3 / 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.

An exemplary description: concrete action verbs, an explicit USE FOR trigger list with synonym coverage, and a DO NOT USE FOR clause that actively routes away sibling-skill conflicts. Its only debatable trait is sheer length (~1000 characters of dense trigger terms), but every term is a functional trigger rather than fluff, so the verbosity guideline is not penalized.

DimensionReasoningScore

Specificity

The description enumerates concrete actions across the full domain — "Build, deploy, evaluate, optimize, fine-tune, and manage Microsoft Foundry agents, models, and resources end to end" plus specific operations like "model fine-tuning (SFT/DPO/RFT)", "dataset curation from traces", and "create Foundry resource". Coverage is comprehensive with no obvious capability gaps, matching the level-5 anchor rather than level 4's "minor gaps in coverage".

5 / 5

Completeness

It explicitly answers both questions: the opening sentence states what the skill does, "USE FOR: foundry, azd ai agent, ... fine-tuning (SFT/DPO/RFT)" gives concrete when-triggers, and "DO NOT USE FOR: Azure Functions, App Service, general Azure deploy (use azure-deploy)" adds explicit negative triggers. This matches level 5's "clearly and explicitly answers both what AND when with concrete trigger phrases"; level 4's weaker "when could be more explicit" does not apply.

5 / 5

Trigger Term Quality

The USE FOR list includes comprehensive natural terms with synonyms — "optimize/improve prompt", "RBAC, role assignment, permissions", "fine-tune, SFT, DPO, RFT", "quota, capacity, region", "deployment failure" — matching level 5's "comprehensive coverage of natural terms including synonyms". It exceeds level 4 ("a few natural terms missing") since both casual ("deploy model", "onboard") and technical phrasings are covered.

5 / 5

Distinctiveness Conflict Risk

The niche is clear (Microsoft Foundry agents/models/resources) with distinct product-specific triggers ("azd ai agent", "Copilot harness", "Agent Optimizer"), and the DO NOT USE FOR clause explicitly disambiguates adjacent skills ("general Azure deploy (use azure-deploy)", "general Azure prep (use azure-prepare)"). This is minimal conflict risk, matching level 5 rather than level 4's "minor overlap risk with closely related skills".

5 / 5

Total

20

/

20

Passed

Validation

93%

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

Validation — 15 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

relative_links

Relative link issues: 63 missing, 60 deeper-than-1-level

Warning

Total

15

/

16

Passed

Repository

microsoft/azure-skills

Path

.github/plugins/azure-skills/skills/microsoft-foundry/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.