Create a blank Copilot Studio agent
Step 1 of 32
Action
Create a blank Copilot Studio agentExpected resultA blank Copilot Studio agent is visible in the captured Draft workspace.
Manual mode · literal browser construction
Build Returns and Complaints Resolution Agent manually.
No PAC CLI, YAML import, or plugin architect. Follow each action and its evidence boundary, compare only the reviewed references, and stop at Draft.
Synthetic disclosure: this is qualitative workflow evidence using packaged synthetic inputs. It is not a customer KPI or a live-system result.
Progress will remain available on this page, but it will not persist after you leave.
Found something inaccurate? Use Report an issue on that step. It opens a prefilled GitHub issue for review and does not submit automatically.
Build and verify
Name Returns and Complaints Manual
Step 2 of 32
Action
Name Returns and Complaints ManualReturns and Complaints Manual
Expected resultThe page header shows the recorded manual build name: Returns and Complaints Manual.
Enter the reviewed global instructions
Step 3 of 32
Action
Enter the reviewed global instructions# Returns and Complaints Resolution Agent — Manual Global Instructions Use only the uploaded anonymous synthetic cases, aggregate trends, safety rules, and operation skills. Do not repeat personal or sensitive free text. Produce review summaries, classifications, draft options, and trend analysis only. Never approve or process a return, refund, credit, replacement, shipment, reservation, account change, or customer message. Lead with case evidence, state the authorized reviewer gate, avoid accusing an individual, and state that no external side effect occurred. The escalated VIP complaint (CMP-5001, synthetic customer David Chen, Diamond VIP, defective ProBook laptop) runs as escalation snapshot, recovery tiers, a Tier 1 plan ready for authorized execution with talking points, a proposed follow-up plan, recovery performance, and an executive summary. Present every dispatch, credit, label, follow-up, and summary as ready for a person to execute or send, never as done. Ten packaged skills cover these steps. <!-- locked-preview-anchors:start --> ## Skill routing map Route from the user's natural-language intent to the correct skill below. Do not narrate internal retrieval, tool selection, restrictions, or implementation mechanics; present only the user-facing result. - `RCR-01` uses skill `anonymous-return-review-queue`. - `RCR-02` uses skill `privacy-safe-complaint-classification`. - `RCR-03` uses skill `human-approved-resolution-options`. - `RCR-04` uses skill `aggregate-returns-quality-trends`. - `RCR-05` uses skill `escalation-snapshot`. - `RCR-06` uses skill `recovery-tiers`. - `RCR-07` uses skill `resolution-execution-plan`. - `RCR-08` uses skill `follow-up-plan`. - `RCR-09` uses skill `recovery-performance`. - `RCR-10` uses skill `executive-summary`. These skill names above are the ONLY valid skill identifiers. Never invent, guess, or reference any other skill name. If the correct skill or its knowledge cannot be loaded after one retry in the same turn, say so honestly and stop. Do not answer using values you already know from these instructions or from general knowledge -- a response with no real citation is not acceptable output. <!-- locked-preview-anchors:end -->
Expected resultThe reviewed manual/GLOBAL-INSTRUCTIONS.md policy is visible or saved without unrecorded edits.
Save the instruction policy
Step 4 of 32
Action
Save the instruction policyExpected resultThe reviewed manual/GLOBAL-INSTRUCTIONS.md policy is visible or saved without unrecorded edits.
Remove default web search
Step 5 of 32
Action
Remove default web searchExpected resultThe captured inventory no longer lists the default web-search capability.
Add knowledge: aibast_returns-complaints-rules-and-safety.md
Step 6 of 32
Action
Add knowledge: aibast_returns-complaints-rules-and-safety.mdExpected resultThe exact knowledge file linked for this step is attached under its complete filename. Wait for processing to finish before continuing. This step does not require the final inventory or prove the other uploads.
Add knowledge: aibast_returns-complaints-synthetic-records.md
Step 7 of 32
Action
Add knowledge: aibast_returns-complaints-synthetic-records.mdExpected resultThe exact knowledge file linked for this step is attached under its complete filename. Wait for processing to finish before continuing. This step does not require the final inventory or prove the other uploads.
Add skill: aibast_complaint-classification_02
Step 8 of 32
Action
Add skill: aibast_complaint-classification_02Expected resultThe linked SKILL.md appears with its frontmatter name and description. Open it and compare the complete instructions with the source file. This step does not require the final inventory or prove the other uploads.
Add skill: aibast_escalation-snapshot_05
Step 9 of 32
Action
Add skill: aibast_escalation-snapshot_05Expected resultThe linked SKILL.md appears with its frontmatter name and description. Open it and compare the complete instructions with the source file. This step does not require the final inventory or prove the other uploads.
Add skill: aibast_executive-summary_10
Step 10 of 32
Action
Add skill: aibast_executive-summary_10Expected resultThe linked SKILL.md appears with its frontmatter name and description. Open it and compare the complete instructions with the source file. This step does not require the final inventory or prove the other uploads.
Add skill: aibast_follow-up-plan_08
Step 11 of 32
Action
Add skill: aibast_follow-up-plan_08Expected resultThe linked SKILL.md appears with its frontmatter name and description. Open it and compare the complete instructions with the source file. This step does not require the final inventory or prove the other uploads.
Add skill: aibast_recovery-performance_09
Step 12 of 32
Action
Add skill: aibast_recovery-performance_09Expected resultThe linked SKILL.md appears with its frontmatter name and description. Open it and compare the complete instructions with the source file. This step does not require the final inventory or prove the other uploads.
Add skill: aibast_recovery-tiers_06
Step 13 of 32
Action
Add skill: aibast_recovery-tiers_06Expected resultThe linked SKILL.md appears with its frontmatter name and description. Open it and compare the complete instructions with the source file. This step does not require the final inventory or prove the other uploads.
Add skill: aibast_resolution-execution-plan_07
Step 14 of 32
Action
Add skill: aibast_resolution-execution-plan_07Expected resultThe linked SKILL.md appears with its frontmatter name and description. Open it and compare the complete instructions with the source file. This step does not require the final inventory or prove the other uploads.
Add skill: aibast_resolution-recommendation_03
Step 15 of 32
Action
Add skill: aibast_resolution-recommendation_03Expected resultThe linked SKILL.md appears with its frontmatter name and description. Open it and compare the complete instructions with the source file. This step does not require the final inventory or prove the other uploads.
Add skill: aibast_return-processing_01
Step 16 of 32
Action
Add skill: aibast_return-processing_01Expected resultThe linked SKILL.md appears with its frontmatter name and description. Open it and compare the complete instructions with the source file. This step does not require the final inventory or prove the other uploads.
Add skill: aibast_trend-analysis_04
Step 17 of 32
Action
Add skill: aibast_trend-analysis_04Expected resultThe linked SKILL.md appears with its frontmatter name and description. Open it and compare the complete instructions with the source file. This step does not require the final inventory or prove the other uploads.
Review model, skills, knowledge, and safety boundaries
Step 18 of 32
Action
Review model, skills, knowledge, and safety boundariesExpected resultThe screenshot visibly confirms Claude Opus 5, 10 rendered skills, and no tools. Knowledge is verified separately in the two knowledge-upload checkpoints.
Review the skill list (10 skills)
Step 19 of 32
Action
Review the skill list (10 skills)Expected resultThe screenshot visibly confirms Claude Opus 5, 10 rendered skills, and no tools. Knowledge is verified separately in the two knowledge-upload checkpoints.
Review the rest of the skill list
Step 20 of 32
Action
Review the rest of the skill listExpected resultThe screenshot visibly confirms Claude Opus 5, 10 rendered skills, and no tools. Knowledge is verified separately in the two knowledge-upload checkpoints.
Open a fresh Preview conversation
Step 21 of 32
Action
Open a fresh Preview conversationExpected resultA fresh Preview surface or its recorded qualitative result is visible; only the evidence file defines a pass.
Pass RCR-01: Customer Service Agent
Step 22 of 32
Action
Pass RCR-01: Customer Service AgentAs Customer Service Agent, summarize the anonymous return review evidence and approval boundary.
Expected resultThe captured Preview evidence records RCR-01 with the recorded identifiers Prepared for:** Customer Service Agent, Draft Return Review Queue, Synthetic size-mismatch case; do not infer results beyond it.
Before this case
Open Preview in a fresh conversation before running RCR-01. A previous response must not influence the evidence.
Pass RCR-02: Customer Service Agent
Step 23 of 32
Action
Pass RCR-02: Customer Service AgentAs Customer Service Agent, classify this product concern without echoing personal information or sending a response.
Expected resultThe captured Preview evidence records RCR-02 with the recorded identifiers Draft Complaint Classification, Product Quality, no return, refund; do not infer results beyond it.
Before this case
Open Preview in a fresh conversation before running RCR-02. A previous response must not influence the evidence.
Pass RCR-03: Customer Service Agent
Step 24 of 32
Action
Pass RCR-03: Customer Service AgentAs Customer Service Agent, draft a policy-grounded option and keep all actions behind authorization.
Expected resultThe captured Preview evidence records RCR-03 with the recorded identifiers Draft Resolution Options, Option for authorized review, no return; do not infer results beyond it.
Before this case
Open Preview in a fresh conversation before running RCR-03. A previous response must not influence the evidence.
Pass RCR-04: Quality Team
Step 25 of 32
Action
Pass RCR-04: Quality TeamAs Quality Team, summarize aggregate defect and return patterns without accusing any person.
Expected resultThe captured Preview evidence records RCR-04 with the recorded identifiers Prepared for:** Quality Team, Synthetic Returns & Complaints Trend Analysis, Key Insights; do not infer results beyond it.
Before this case
Open Preview in a fresh conversation before running RCR-04. A previous response must not influence the evidence.
Pass RCR-05: Service Manager
Step 26 of 32
Action
Pass RCR-05: Service ManagerAs Service Manager, I have an escalated customer complaint about a defective laptop and the customer is very upset; help me resolve it quickly.
Expected resultThe captured Preview evidence records RCR-05 with the recorded identifiers David Chen, $18,400, Churn Risk:** 87%; do not infer results beyond it.
Before this case
Open Preview in a fresh conversation before running RCR-05. A previous response must not influence the evidence.
Pass RCR-06: Service Manager
Step 27 of 32
Action
Pass RCR-06: Service ManagerAs Service Manager, show me resolution options that will keep this VIP customer satisfied.
Expected resultThe captured Preview evidence records RCR-06 with the recorded identifiers Premium Recovery, $540, 34:1 ROI; do not infer results beyond it.
Before this case
Open Preview in a fresh conversation before running RCR-06. A previous response must not influence the evidence.
Pass RCR-07: Service Manager
Step 28 of 32
Action
Pass RCR-07: Service ManagerAs Service Manager, Tier 1 is approved; get it ready and prepare my talking points for the call.
Expected resultThe captured Preview evidence records RCR-07 with the recorded identifiers ready for authorized execution, ProBook Elite Plus, Talking Points; do not infer results beyond it.
Before this case
Open Preview in a fresh conversation before running RCR-07. A previous response must not influence the evidence.
Pass RCR-08: Service Manager
Step 29 of 32
Action
Pass RCR-08: Service ManagerAs Service Manager, the call went well and the customer accepted; what is our follow-up plan?
Expected resultThe captured Preview evidence records RCR-08 with the recorded identifiers Day 30, delivery confirmation SMS, Nothing has been scheduled; do not infer results beyond it.
Before this case
Open Preview in a fresh conversation before running RCR-08. A previous response must not influence the evidence.
Pass RCR-09: Service Manager
Step 30 of 32
Action
Pass RCR-09: Service ManagerAs Service Manager, show me the service recovery performance and financial impact.
Expected resultThe captured Preview evidence records RCR-09 with the recorded identifiers 4.2 hours, $127,000, $4.67M; do not infer results beyond it.
Before this case
Open Preview in a fresh conversation before running RCR-09. A previous response must not influence the evidence.
Pass RCR-10: Service Manager
Step 31 of 32
Action
Pass RCR-10: Service ManagerAs Service Manager, create the executive summary and recap what we accomplished.
Expected resultThe captured Preview evidence records RCR-10 with the recorded identifiers Executive Summary, 87% churn risk, Business Value; do not infer results beyond it.
Before this case
Open Preview in a fresh conversation before running RCR-10. A previous response must not influence the evidence.
Confirm Draft and stop before publish
Step 32 of 32
Action
Confirm Draft and stop before publishExpected resultThe agent remains Draft and no Publish action is taken.
Troubleshooting
A screenshot or browserfilm frame is missing
Do not invent, recreate, or substitute an image. Keep the step open until its actual evidence is captured and reviewed. Remove --allow-pending only after all workshop evidence gates are met.
A knowledge file is still processing
Wait for ingestion to finish before Preview. A partial answer is not evidence.
A skill upload fails
Use the linked raw SKILL.md. Fix the reviewed source deliberately; do not silently skip the action.
The model differs from Easy mode
Record the substitution and stop the parity claim until Easy and Manual use the same reviewed model.
The Preview answer misses an identifier
Mark the recorded case failed, inspect instructions and inventory, then replay the exact prompt in a fresh conversation.
Should I publish?
No. Keep the same owned manual agent in Draft. Do not choose Publish as part of this tutorial.