Understand your request
Identify the client, the result, and the format. Ask about an important missing detail only if it is not already saved.
Name the client, say what you want made, and tell me what is different this time. You do not need to remember folders or special commands.
“Write one 40-second podcast ad for this client. Use the saved research and current rules. Give me the script and a short brief.”
Identify the client, the result, and the format. Ask about an important missing detail only if it is not already saved.
Use the saved steps for this kind of job. Writing a script, making a video, and doing research have different starting points.
Load their product details, research, restrictions, and relevant feedback. Keep one client's rules with that client.
Read the short method for this task. Open the original video or source passage when a detail needs checking.
Use the agreed structure and this client's evidence. Produce the amount and format you asked for.
Check the actual result, save it in the usual place, and give you the link or file. Save lasting corrections for the next task.
Read left to right, then continue on the second row. Feedback returns to the client files before the next task.
Choose a situation to see what you say and what I do.
You can give me a file or link and ask me to save it. These are the places I will use.
What they sell, who they serve, and where their detailed instructions live.
clients/{client}/PROFILE.yamlWhat should change next time, which tasks it applies to, and how to check it.
clients/{client}/FEEDBACK.yamlThis client's reviews, interviews, customer language and research findings.
clients/{client}/ads/research/Original sources stay in research. Short, reviewed methods live in the shared guides.
research/learnings/ → skills/knowledge/A task list points to the right workflow, client inputs, output format and checks.
directives/core/workflows.yaml → skills/agent/A short note records decisions, saved work and the next step so we can resume.
workspace/tasks/{run}/STATE.mdUnsure where a file goes? Say “save this for this client” or “save this as shared learning.” If you prefer to drop files yourself, use research/inbox/ or a client's inbox/. I will file them and tell you where they went.
For a new client: their product, audience, offer, brand files, restrictions and useful past work. Share what you have; I will identify what is missing for the job.
For each task: the result, quantity, format, and anything different this time. You do not need to repeat information that is already saved.
The same starting files, a clear method, the required output format, and checks before delivery. Feedback is saved with the right client and task.
Creative wording can vary. The system helps catch mistakes and prevent repeats; it cannot promise perfect judgment or results every time.
20 videos downloaded, 20 timestamped transcripts saved, and screen samples inspected from every video. The practical ideas were compared with current official documentation and the existing workspace.
Review covered full transcripts or relevant sections and sampled screens, rather than continuous viewing of all footage. Some linked templates were unavailable without signup or purchase.
Rollout status: 18 task routes are installed. 86 client folders have readable profiles and a place for feedback. Two client source maps have been reviewed in detail. The remaining maps are marked for review; missing client facts are left unknown. This is not a claim that every historical file or client deliverable has been checked.
Technical reference: OpenAI's project instructions, OpenAI's skills documentation, and T3's source. The detailed internal evidence report is saved in the workspace.