Course documentation

Delivery and format

Available live (instructor-led) or asynchronously (self-paced), and virtually or in person. A participant enrolling individually selects the delivery mode at registration; where an organization sponsors a cohort, the organization selects the combination that fits its people, and modes may be mixed within a single cohort. Learning objectives, required practical deliverables, and assessment standards are identical across every delivery mode.

Instructor-led session with hands-on practice blocks. Live delivery runs as a single 3-hour block.

• 3 hours of instruction — live instructor-led, or the equivalent asynchronous curriculum
• 3 hours of role-customized self-paced curriculum delivered through The SkillSpout Current, SkillSpout's client learning platform, available from 30 days before the course through 30 days after

6 clock hours per participant. This course includes no coaching session; individual coaching begins at AI at Work.

Course content covers the major platforms comparatively — Microsoft Copilot, Claude, ChatGPT, and Google Gemini — rather than training a single vendor's product, so that participants can operate whichever platform they work in.

No hard limit; sized to preserve hands-on facilitation.

Before the course

Each participant names a real, recurring task from their current role that they want AI to take off their plate. That task is held open through the session and solved by the participant in the final module.

Learning objectives

1. Select the appropriate AI system type for a given work task — traditional, generative, or agentic — route the task to the tool that fits it, and distinguish work that belongs to deterministic automation rather than to a language model.
2. Account for why an AI tool underperforms on a given task, distinguishing a limit of the model from a matter of grounding, configuration, or the data the tool can reach.
3. Select an appropriate model for a task by speed, complexity, and cost; manage the context window as working memory; and organize work across projects and sessions, including the failure modes of standing custom instructions.
4. Construct a structured prompt using three required moves — be specific, give context, define the output — and refine it iteratively, including requesting sources, requesting a stated confidence level, and directing the model to improve its own instruction.
5. Prepare source material an AI can read reliably, selecting formats that survive ingestion and transcribing audio before submission, and identify where a connector is required and what approval it needs.
6. Configure the security and privacy settings of an AI platform account, including multi-factor authentication by authenticator application, disabling chat history for sensitive work, opting out of training-data use, and separating personal from organizational accounts.
7. Detect fabricated citations, inferred private information, scope drift, and bias in AI-generated output, recognize unwarranted agreement as a failure mode rather than confirmation, and correct or reject the output accordingly.
8. Identify a prompt-injection attempt in material an AI tool reads on the user's behalf, and apply the controls that contain it.
9. Adversarially review AI output through a second model, and apply a four-question verification checkpoint before any AI-assisted output is released, withholding the classes of information that may never be entered into an AI tool.
10. Evaluate two AI platforms against the same task and judge the results on whether each is accurate, grounded in real source material, and usable as delivered.
11. Produce a working AI-assisted solution to a recurring task from their own job, and quantify the time the task takes without AI, with AI, and per year.

Module outline

Module outline (3-hour block — live sequence; the asynchronous curriculum covers the same modules and objectives)

• Welcome and objectives. The session is built from the pre-course survey. Each participant names a real recurring task for AI to take off their plate; held open until the final module
• The landscape: three types of AI, and what the generative kind is actually doing. It predicts language, and looks up only what it is connected to
• Why a tool underperforms: separating a limit of the model from a matter of grounding, configuration, and what the tool can reach, using the participants' own experience as the evidence
• Generative AI versus deterministic automation — matching the task to the right tool, and naming the work that never goes to a language model
• Model, context and memory: selecting a model by task, speed and cost; the context window as working memory; projects versus fresh sessions; and the failure modes of standing custom instructions
• Break
• The three moves: be specific, give context, define the output, then iterate. Live construction from the front, with the room filling the brackets
• Prompt lab (hands-on) — each participant rebuilds a prompt for work they did this week, then applies the follow-up moves: ask for sources, request a stated confidence level, and direct the model to improve its own instruction
• Source material an AI can read: formats that survive ingestion, transcription before submission, and connectors — what they are and the approval path they require
• Verification and responsible use (hands-on) — fabricated citations and confident wrong answers; unwarranted agreement as a failure mode; bias; data exposure and the boundary between personal and organizational accounts; prompt injection in material the tool reads on your behalf; adversarial review through a second model; and a safe-or-not judgement exercise
• Security and privacy configuration walkthrough — multi-factor authentication by authenticator application, chat history, training-data opt-out, feedback controls as a consent signal, and the interim data rule that applies until a written policy exists
• The platform decision (hands-on) — the same task run head to head on two platforms, judged by the room on three criteria: accurate, grounded, and usable as delivered. Where platforms differ in practice, including image generation and text extraction
• The reframe: the platform matters less than where the organization's data lives and what the tool can reach
• Applied task (hands-on) — each participant solves the task named in the opening module, applies the four-question release checkpoint to the output, and quantifies the time recovered
• Platform onboarding — guided navigation of The SkillSpout Current toward one stated goal; each participant queues a specific first item
• Written commitment — the named task, the named day, and the minutes it takes today, recorded for the post-course assessment

Assessment

• Pre-course and post-course skills and confidence assessment, administered to every participant. Results are reported to the participant, and to the sponsoring organization where one has enrolled the participant.
• Practical assessment: a working AI-assisted solution to the participant's own named task, with the time recovered quantified, and a completed release checkpoint applied to that output.

What the participant receives

• Certificate of completion, issued per participant, showing participant name, course name, and date completed.

Delivery and format

Available live (instructor-led) or asynchronously (self-paced), and virtually or in person. A participant enrolling individually selects the delivery mode at registration; where an organization sponsors a cohort, the organization selects the combination that fits its people, and modes may be mixed within a single cohort. Learning objectives, required practical deliverables, and assessment standards are identical across every delivery mode.

Hands-on lab. Live delivery runs as a single 4-hour block on one day, or split into two 2-hour sessions.

• 4 hours of instruction — live instructor-led, or the equivalent asynchronous curriculum
• Two 30-minute one-on-one coaching sessions with an instructor, applied to the participant's own work, scheduled virtually or in person
• 4 hours of role-customized self-paced curriculum, delivered through The SkillSpout Current, available from 30 days before the course through 30 days after
• 3 hours of practical project work producing the required deliverable

12 clock hours per participant.

Delivered in the AI platform the participant works in, whether licensed by their organization or their own account — Microsoft Copilot, Claude, ChatGPT/Custom GPTs, or Google Gemini. Course materials are customized to the participant's actual tool stack before delivery. Exercises run against the participant's own platform, not a demonstration environment.

Minimum 8 participants for private delivery.

Before the course

Each participant completes a survey identifying a real, recurring task from their current role. That task becomes the participant's working problem for the course.

Learning objectives

1. Evaluate a recurring work task and determine whether it is suited to AI execution, partial AI assistance, or neither.
2. Construct a structured prompt with an advanced prompting framework, and refine it iteratively against output.
3. Configure refusal behavior and required citations in a prompt so that high-stakes output carries a verifiable audit trail.
4. Operate the AI assistants built into their own productivity suite — Microsoft 365 with Copilot, Google Workspace with Gemini, or the equivalent — to complete real work in the mail client, the document editor and the presentation tool: locating the assistant's entry point across the versions of each application in use, supplying context, and refining output both by preset option and by direct instruction.
5. Account for the operating limits of an in-application assistant, including its shorter working memory relative to the suite's chat interface, and how the suite indexes stored files, what delays retrieval of a newly added document, and how to force a reindex.
6. Operate the assistants a suite provides pre-built, including directing a prompt-refinement assistant to interrogate a request before drafting against it.
7. Produce three tested use cases drawn from the participant's own job, and build a prompt library from them that can be shared with colleagues.
8. Document a completed AI-assisted process in sufficient detail that a colleague can execute it, treating every AI output as a first draft rather than a final one.

Module outline

Module outline (4 hours — live sequence; the asynchronous curriculum covers the same modules and objectives)

• Welcome and prior-course recap; wins and blockers
• Applying AI to existing expertise: selecting the tasks in each participant's role where it pays
• Advanced prompting lab (hands-on) — structured framework, live experimentation, refusal behavior and citation requirements for high-stakes work, rewrite-and-improve exercise
• Workflow adoption: each participant commits to one recurring workflow as their working project
• Break
• Where language models fail, and the verification step that catches each failure
• The AI already in the participant's productivity suite — the assistants the suite provides pre-built and what each is for; the assistant's entry point in the mail client, the document editor and the presentation tool, and how that differs across the versions of each application in use; supplying context; refining by preset option and by direct instruction; the shorter working memory of an in-application assistant; and how the suite indexes stored files, what delays retrieval, and how to force a reindex
• Break
• Tool-stack application lab (hands-on) — extended work in the participant's own suite on their adopted workflow, using a prompt-refinement assistant to interrogate the request before drafting; each participant lands three tested use cases and starts a prompt library that can be shared with colleagues
• Commitments, documentation, and coaching scheduling

Assessment

• Pre-course and post-course skills and confidence assessment, administered to every participant. Results are reported to the participant, and to the sponsoring organization where one has enrolled the participant.
• Practical assessment: three tested use cases and a personal prompt library, produced from the participant's own work, plus a documented process a colleague can execute.

What the participant receives

• Certificate of completion, issued per participant, showing participant name, course name, and date completed.

Delivery and format

Available live (instructor-led) or asynchronously (self-paced), and virtually or in person. A participant enrolling individually selects the delivery mode at registration; where an organization sponsors a cohort, the organization selects the combination that fits its people, and modes may be mixed within a single cohort. Learning objectives, required practical deliverables, and assessment standards are identical across every delivery mode.

Hands-on build session, run on a design-thinking model rather than lecture. Live delivery runs as a single 4-hour block on one day, or split into two 2-hour sessions.

• 4 hours of instruction — live instructor-led, or the equivalent asynchronous curriculum
• Two 30-minute one-on-one coaching sessions with an instructor, applied to the participant's own build, scheduled virtually or in person
• 4 hours of role-customized self-paced curriculum, delivered through The SkillSpout Current, available from 30 days before the course through 30 days after
• 3 hours of practical project work producing the required deliverable

12 clock hours per participant.

Minimum 8 participants.

Learning objectives

1. Determine when a custom AI assistant is the correct tool, as opposed to a single prompt or a specialist chat session, and scope a build whose value justifies its cost.
2. Configure all three components of an AI assistant — model, instructions, and user experience — and build its instruction set across all three layers: behavior, boundaries, and data.
3. Write effective custom instructions (Layer 1, behavior) with the full anatomy: role and persona, the audience it writes for, tone and style, scope and boundaries, and output format.
4. Design guardrails and refusal-behavior rules (Layer 2, boundaries) calibrated to the stakes and audience of the assistant.
5. Curate and integrate knowledge sources (Layer 3, data) — folder-level rather than file-level, in supported storage, within context limits — in compliance with the applicable data classification policy.
6. Build, test, and troubleshoot a working assistant that applies all three layers to a real departmental task.
7. Assign and document maintenance ownership for any assistant shared beyond its builder.

Module outline

Module outline (4 hours — live sequence; the asynchronous curriculum covers the same modules and objectives)

• Builder framing: building for teams, not only for self
• Project share-out: what each builder shipped and where it stalled; stalls become build candidates
• Declare your build; scope triage; when to build versus when not to
• Anatomy of an assistant; the three instruction layers, applied to each builder's declared build
• Layer 1 — custom instructions: teach, then individual drafting of the five fields, then pair review and debrief
• Layer 2 — guardrails and refusal behavior: writing refusal rules calibrated to stakes; direct-to-human and cite-source rules
• Prompt refinement step, run against the platform's refiner
• Break
• Layer 3 — knowledge integration: folders not files, supported storage, curation and context limits, applied against the applicable data classification policy
• Build your assistant (hands-on) — configuration in the participant's platform: refined instructions, knowledge folder, capability settings
• Test and troubleshoot against real tasks
• Share-out: each builder demonstrates their assistant
• Assistant management, permissions, and maintenance ownership

Assessment

• Pre-course and post-course skills and confidence assessment, administered to every participant. Results are reported to the participant, and to the sponsoring organization where one has enrolled the participant.
• Practical assessment: each participant must produce a working assistant applying all three instruction layers, demonstrate it, and name its maintenance owner.

What the participant receives

• Certificate of completion, issued per participant, showing participant name, course name, and date completed.

Delivery and format

Available live (instructor-led) or asynchronously (self-paced), and virtually or in person. A participant enrolling individually selects the delivery mode at registration; where an organization sponsors a cohort, the organization selects the combination that fits its people, and modes may be mixed within a single cohort. Learning objectives, required practical deliverables, and assessment standards are identical across every delivery mode.

Hands-on lab. Live delivery runs as a single 4-hour block on one day, or split into two 2-hour sessions.

• 4 hours of instruction — live instructor-led, or the equivalent asynchronous curriculum
• Two 30-minute one-on-one coaching sessions with an instructor, applied to the participant's own configuration, scheduled virtually or in person
• 4 hours of role-customized self-paced curriculum, delivered through The SkillSpout Current, available from 30 days before the course through 30 days after
• 3 hours of practical project work, including assembling the participant's own source material

12 clock hours per participant.

Delivered on the agentic work platform the organization uses. SkillSpout's materials exist for Claude Cowork. Equivalent environments are supported at the organization's election, including Microsoft Copilot Cowork, ChatGPT Work, and Gemini Spark. Learning objectives, the required deliverable and the assessment standard are identical on any supported platform. Where the platforms differ in a way that matters to the work, the course teaches the difference rather than training a single vendor's product.

Learning objectives

1. Operate an agentic work environment: sessions, connectors, the skill panel, and file delivery.
2. Select between a chat session, a configured assistant, and an agentic system for a given task, and route the work accordingly.
3. Configure a connected knowledge folder as the agent's source of voice, brand, examples, and rules, and edit it so changes take effect without republishing.
4. Assemble that knowledge folder from real work material, filling every voice and brand file from genuine source content rather than allowing the agent to generate it.
5. Apply command namespacing to a distributable plugin so that it never occupies a generic command name.
6. Configure per-person permissions on a shared knowledge folder through the storage platform's own permission model, setting team read access and individual edit access on personal voice files.
7. Verify that the placeholder gate holds — that the agent halts and asks rather than inventing content when a required field is unfilled.
8. Run an end-to-end agentic task against real work material and verify the output against its source.

Module outline

Module outline (4 hours — live sequence; the asynchronous curriculum covers the same modules and objectives)

• Welcome, norms, parking lot, agenda
• The engine and the knowledge folder: what SkillSpout builds and versions, what the participant's organization owns and edits; guided tour of an installed plugin
• Routing lab (hands-on) — sorting real work tasks across chat, configured assistant, and agentic system against the one-question-versus-multi-step rule
• Connecting the knowledge folder (hands-on) — pointing the agent at a whole folder rather than individual files, and confirming subfolder inclusion
• Break
• Assembling the knowledge folder (hands-on) — filling voice, brand, example, and rule files from real work material
• The placeholder gate (hands-on) — triggering it deliberately, and why the agent must never generate a brand voice
• Wrap; each participant names the source material they will supply before session 2
• Homework check: what landed in each knowledge folder, and where it stalled
• Production home (hands-on) — shared cloud folder mounted locally, one shared source, no drift
• Per-person permissions (hands-on) — team read on the whole folder, individual edit on each person's own voice file; testing the boundary
• Command namespacing — applying the organization prefix convention to a distributable plugin
• Break
• End-to-end agentic task (hands-on) — run against the participant's own real material, output verified against source
• Maintenance ownership: who edits the knowledge folder, and what changes take effect without a republish
• Commitments and coaching scheduling

Assessment

• Pre-course and post-course skills and confidence assessment, administered to every participant. Results are reported to the participant, and to the sponsoring organization where one has enrolled the participant.
• Practical assessment: each participant must produce a configured knowledge folder in production storage with per-person permissions set, and run one end-to-end agentic task against real material with the output verified against its source.

What the participant receives

• Certificate of completion, issued per participant, showing participant name, course name, and date completed.

Delivery and format

Available live (instructor-led) or asynchronously (self-paced), and virtually or in person. A participant enrolling individually selects the delivery mode at registration; where an organization sponsors a cohort, the organization selects the combination that fits its people, and modes may be mixed within a single cohort. Learning objectives, required practical deliverables, and assessment standards are identical across every delivery mode.

Hands-on lab. Live delivery runs as a single 4-hour block on one day, or split into two 2-hour sessions.

• 4 hours of instruction — live instructor-led, or the equivalent asynchronous curriculum
• Two 30-minute one-on-one coaching sessions with an instructor, applied to the participant's own build, scheduled virtually or in person
• 4 hours of role-customized self-paced curriculum, delivered through The SkillSpout Current, available from 30 days before the course through 30 days after
• 3 hours of practical project work. The administrator setup session is a prerequisite to the Agentic series and is not counted here

12 clock hours per participant.

Delivered on the agentic work platform the organization uses. SkillSpout's materials exist for Claude Cowork. Equivalent environments are supported at the organization's election, including Microsoft Copilot Cowork, ChatGPT Work, and Gemini Spark. Learning objectives, the required deliverable and the assessment standard are identical on any supported platform. Where the platforms differ in a way that matters to the work, the course teaches the difference rather than training a single vendor's product.

Learning objectives

1. Select a candidate task from their own work against three suitability patterns — multi-step, operates on real files, or spans multiple tools — and score it against the alternatives.
2. Configure a dedicated working folder and a structured source, action, output, and bounds specification for an agentic task, and run it.
3. Write the three parts of a skill — name, description, and instructions — so that the description routes reliably.
4. Design a skill before building it, from a four-quadrant design canvas: the task, the definition of done, the required inputs, and the hard limits the skill must never exceed.
5. Apply the operating safety rules: skills execute with the user's own permissions; hard limits belong in the instructions; sensitive data is never stored in a skill.
6. Build, test, and run a working reusable skill, and package and share it with a colleague as a distributable file.

Module outline

Module outline (4 hours — live sequence; the asynchronous curriculum covers the same modules and objectives)

• Welcome, norms, parking lot, agenda
• Routing check: sorting each participant's brought tasks across chat, configured assistant, and agentic system
• Candidate identification: the recurring work each participant should have delegated
• Connector verification (full configuration completed in pre-work)
• Task-suitability patterns; participants score and select their own candidate
• Safe and secure operation: built-in safeguards, three operating habits, mapping risk to safeguard
• Demonstration: a live end-to-end agentic task
• Hands-on: first agentic task — dedicated working folder, structured source/action/output/bounds specification, run against the participant's own selected task
• Bridge: how repeated prompts and scheduled tasks become reusable skills
• Break
• Anatomy of a skill: name, description, instructions
• Each participant names one repeatable task to build
• Safety, scope, and boundaries: permissions, draft-versus-send precision, one task per skill, never store credentials
• Skill design canvas — guided work across all four quadrants
• Build (hands-on) — heads-down build from the canvas, with facilitator circulation
• Test and deploy — test against real work data, grade output, iterate, validate guardrails, package for sharing
• Wrap and coaching scheduling

Assessment

• Pre-course and post-course skills and confidence assessment, administered to every participant. Results are reported to the participant, and to the sponsoring organization where one has enrolled the participant.
• Practical assessment: each participant must produce a working installed skill, a completed design canvas, and a named implementation commitment.

What the participant receives

• Certificate of completion, issued per participant, showing participant name, course name, and date completed.

Delivery and format

Available live (instructor-led) or asynchronously (self-paced), and virtually or in person. A participant enrolling individually selects the delivery mode at registration; where an organization sponsors a cohort, the organization selects the combination that fits its people, and modes may be mixed within a single cohort. Learning objectives, required practical deliverables, and assessment standards are identical across every delivery mode.

Hands-on build session, chaining the skills built in AI Agent Skill Development into multi-step systems and operating the agents that run them. Live delivery runs as a single 4-hour block on one day, or split into two 2-hour sessions. Where the session is split, the live run is launched in the first session and its manifest is read and corrected in the second.

• 4 hours of instruction — live instructor-led, or the equivalent asynchronous curriculum
• Two 30-minute one-on-one coaching sessions with an instructor, applied to the participant's own build, scheduled virtually or in person
• 4 hours of role-customized self-paced curriculum, delivered through The SkillSpout Current, available from 30 days before the course through 30 days after
• 3 hours of practical project work, including advance selection and preparation of safe test data

12 clock hours per participant.

Delivered on the agentic work platform the organization uses. SkillSpout's materials exist for Claude Cowork. Equivalent environments are supported at the organization's election, including Microsoft Copilot Cowork, ChatGPT Work, and Gemini Spark. Learning objectives, the required deliverable and the assessment standard are identical on any supported platform. Where the platforms differ in a way that matters to the work, the course teaches the difference rather than training a single vendor's product.

Learning objectives

1. Determine whether a workflow from their own job is an agentic candidate — requiring judgment, multi-step coordination, or multi-source assembly — or is better served by simpler automation.
2. Decompose that workflow into named single-purpose components with defined handoffs, dependencies, and human approval gates.
3. Build an orchestrator, single-purpose workers, and an independent verifier, and place each human approval gate by cost of error.
4. Apply five reliability constraints when building: an explicit allowlist of callable components rather than open discovery; a single input gate at the front; input ownership held by the orchestrator rather than the workers; halt-on-failure rather than degraded continuation; and adversarial review of the final step.
5. Build a working multi-step system from a design specified on paper, and launch a live run against real test data.
6. Read a run manifest, apply a six-point test protocol, separate a component fault from a design fault, and correct the single weakest link.

Module outline

Module outline (4 hours — live sequence; the asynchronous curriculum covers the same modules and objectives)

• Act I — See it work. The architectural limit of a single overloaded component; anatomy of a chain (orchestrator, workers, verifier); demonstration of a working chain executing four jobs from one command; each participant names their target workflow
• Fit check. Each target workflow is run through a suitability filter; participants confirm a chain candidate, park simpler automations, and select a backup
• Act II — Design it on paper. Chain overview map and one component card per box: handoffs, dependencies, human gates. Paired review, in which the reviewer's task is to find any component doing two jobs
• Act III(a) — The build specification. Participants write the full specification from their design rather than starting with a component generator
• Act III(b) — Build and launch (hands-on). Minimum build: orchestrator, two workers, verifier. Each participant launches their system's first live run before the break
• Break — systems execute; facilitator triages launch failures
• Act III(c) — Instruction while systems run. The five reliability rules; placing human gates by cost of error; manifest anatomy; common build failures and their signatures
• Act IV — Verify and troubleshoot (hands-on). Read the manifest, walk the six-point test protocol, separate component faults from design faults, apply one correction, re-run the weakest link
• Commitments and wrap; commitments become the coaching agenda

Assessment

• Pre-course and post-course skills and confidence assessment, administered to every participant. Results are reported to the participant, and to the sponsoring organization where one has enrolled the participant.
• Practical assessment: each participant must produce a multi-step system that has actually executed during the session, with launch evidence in the run manifest, at least one correction applied, and completed design artifacts.

What the participant receives

• Certificate of completion, issued per participant, showing participant name, course name, and date completed.

Delivery and format

25 clock hours.

Learning objectives

1. Deliver each of the three Generative series courses to a group of learners, following the published run of show and adapting timing to the room.
2. Facilitate the hands-on lab segments, circulating among builders, diagnosing failures in a participant's build, and correcting them without taking over the work.
3. Conduct one-on-one coaching sessions on a participant's own build, including diagnosing whether a stalled build is a component fault or a design fault.
4. Customize course materials to the licensed AI platform and tool stack of the organization being taught and actual tool stack without altering the learning objectives or the assessment standard.
5. Apply the applicable data classification policy when instructing participants on what may and may not be used with an AI tool.
6. Administer pre- and post-training assessment, and configure and report the adoption measurement a certified instructor is accountable for.
7. Evaluate a participant's practical deliverable against the course's completion standard.

Components

• Trainer preparation — run-of-show, customization rules, and assessment standards for three courses — 13 hours
• Practice delivery — rehearsed delivery of course segments against the published run of show — 6 hours
• Certification examination, written — course material, its rationale, and the assessment standards — 2 hours
• Certification examination, performance — delivery of assigned course material, submitted as a recording and scored against a published rubric — 4 hours

Assessment

• Certification is granted only after the candidate has passed both parts of the certification examination: the written examination and the assessor-scored performance assessment. Assessment is individual; there is no cohort pass.

What the participant receives

• Certificate of certification, issued per candidate, showing candidate name, credential name, and date earned. Credentials issue from SkillSpout.

Delivery and format

25 clock hours.

Learning objectives

1. Deliver all three Agentic series courses to a group of learners, following the published run of show and adapting timing to the room.
2. Deliver on either supported agentic platform, customizing materials to whichever platform is in use without altering the learning objectives or the assessment standard.
3. Facilitate a room in which participants' multi-step systems are executing live, including triaging launch failures within the session.
4. Read a run manifest on a participant's build, apply the six-point test protocol, and separate a component fault from a design fault under live conditions.
5. Conduct one-on-one coaching on a participant's agentic build, diagnosing whether a stalled build is a component fault or a design fault.
6. Instruct participants in the operating safety rules and verify their application: permissions, hard limits in instructions, and never storing sensitive data in a skill.
7. Configure and report the output-multiple adoption measurement a certified agentic instructor is accountable for.
8. Evaluate a participant's practical deliverable against the course's completion standard.

Components

• Trainer preparation — run-of-show, customization rules, and assessment standards for three courses, across both supported agentic platforms — 11 hours
• Practice delivery — rehearsed delivery of course segments, including in-session triage of launch failures — 6 hours
• Certification examination, written — course material, platform differences, and the reliability rules — 3 hours
• Certification examination, performance — delivery of assigned course material plus a live troubleshooting scenario on another person's build, submitted as a recording and scored against a published rubric — 5 hours

Assessment

• Certification is granted only after the candidate has passed both parts of the certification examination: the written examination, and the assessor-scored performance assessment, which includes the live troubleshooting scenario. Assessment is individual; there is no cohort pass.

What the participant receives

• Certificate of certification, issued per candidate, showing candidate name, credential name, and date earned. Credentials issue from SkillSpout.