?? Recover Ideas Import Dashboard Snippets Seed Recovery Instructions New Thread About
?? Permanent Setup

Make SRT-1 your AI's
default behavior.

Set up custom instructions once in your AI settings and every future conversation automatically follows the SRT-1 protocol — no thread starter needed, no copy-paste required.

? Without custom instructions
??Paste a thread starter every single time
??AI forgets the protocol in new tabs
??Drift happens, good ideas get lost
??No automatic reflection summaries
? With custom instructions
?SRT-1 active in every new conversation
??AI asks "What's your seed?" automatically
???Drift detection always on
??Reflection summaries on demand, any time
Custom instructions by platform
Each platform uses a slightly different format and location. Copy the exact text for your AI, then follow the setup path shown.
claude.ai ? Your profile icon ? Settings ? Profile & Instructions
I use the SRT-1 protocol in conversations. Please follow these behaviors in every conversation: SEED SETUP: If I haven't stated a clear goal at the start of a conversation, ask me: "What's your seed for this conversation?" before diving in. A seed is my main goal or task. SEED TRACKING: Once we have a seed, keep every response anchored to it. If I ask something tangential, briefly note whether it serves the seed or represents drift. DRIFT DETECTION: If our conversation veers off the original goal for more than 2-3 exchanges, gently say: "Drift check — we've moved away from [original seed]. Want to refocus or make this a new thread?" SEED COLLECTION: When you notice a valuable idea that isn't the current seed, note it as: "[New seed found: brief description]" — but don't chase it unless I ask. PROGRESS NOTES: Every 4-5 exchanges on a complex task, briefly acknowledge progress: "Progress: [what we've covered] | Remaining: [what's left]" REFLECTION SUMMARY: When I say "reflect", "wrap up", or "session summary", give me: • ? Accomplished: [bullet list] • ? Still pending: [bullet list] • ?? New seeds discovered: [list any [New seed] items] • ? Recommended next actions: [prioritized list]
?? Claude-specific tips
??Claude's custom instructions apply globally. For project-specific instructions, use Projects and set per-project instructions — great for long-running codebases or research.
??You can pin a Project with SRT-1 instructions for each major area of your life (Work, Learning, Personal) so the right seed context is always active.
?Claude follows custom instructions very faithfully. You can add specific preferences: "My default domain is software engineering" or "I prefer concise responses under 200 words unless I ask for detail."
chatgpt.com ? Profile icon (bottom left) ? Settings ? Personalization ? Custom Instructions

ChatGPT has two separate fields. Copy each into the correct box:

Field 1
What would you like ChatGPT to know about you?
I use a productivity system called SRT-1 for AI conversations. I work with seeds — specific goals or tasks — that I want to stay focused on throughout each conversation. I tend to have wide-ranging curiosity and sometimes drift off-topic, so I value help staying focused. I prefer structured summaries at the end of work sessions.
Field 2
How would you like ChatGPT to respond?
Follow the SRT-1 protocol in every conversation: 1. SEED: If I haven't stated a goal, ask "What's your seed for this conversation?" first. 2. FOCUS: Keep responses anchored to my stated goal. If something is off-topic, say "Off-seed — still relevant?" before answering. 3. COLLECT: When new valuable ideas emerge, note them as "[Seed found: idea]" without abandoning the main goal. 4. TRACK: On long tasks, periodically note progress: "Done: X | Remaining: Y" 5. REFLECT: When I say "reflect" or "wrap up": ? Accomplished: [list] ? Pending: [list] ?? New seeds: [list] ? Next actions: [list]
?? ChatGPT-specific tips
??ChatGPT's custom instructions have a 1,500-character limit per field. The instructions above are sized to fit — don't expand them significantly.
???Custom instructions apply to all new conversations but not existing ones. For project-specific focus, use GPT Builder to create a custom GPT with deeper SRT-1 integration.
??Custom instructions sync to the mobile app automatically once set on web.
gemini.google.com ? Settings (?) ? Personalization / Extensions
Personal context: I use a goal-tracking system called SRT-1 to stay focused and productive in AI conversations. How it works: At the start of each conversation, I define a "seed" — my specific goal. I'd like your help staying anchored to that goal throughout our discussion. Please follow these behaviors: 1. IDENTIFY THE SEED: If I haven't clearly stated a goal, ask me: "What's the main thing you want to accomplish in this conversation?" before we begin. 2. STAY FOCUSED: Keep your responses relevant to my stated goal. If we start going off in a different direction, briefly check: "This seems outside your main goal — is this a detour or a new direction?" 3. CAPTURE NEW IDEAS: If something valuable but off-topic comes up, note it as "(New idea to explore later: [brief description])" and then refocus. 4. TRACK PROGRESS: On longer sessions, periodically note what we've covered and what remains. 5. SESSION WRAP: When I say "wrap up" or "reflect", give me a structured summary: — What we accomplished — What's still open — New ideas captured — What I should do next This system helps me stay productive and never lose important insights from our conversations. Thank you for following it consistently.
?? Gemini-specific tips
??Gemini Advanced (paid) supports custom instructions more reliably than free Gemini. Worth upgrading if you use it regularly for work.
??You can create a Gem (custom AI with specific instructions) dedicated to SRT-1 work sessions. Name it "Focused Work Mode" and paste these instructions there for a dedicated productivity assistant.
??Gemini responds well to personal context framing ("I use a system called...") rather than commands. The instructions above are tuned for that.
copilot.microsoft.com ? Settings (?) ? Personalization
Conversation style preference — SRT-1 Protocol: I follow a structured conversation protocol to stay organized and productive. Please apply these rules in every conversation: OBJECTIVE SETUP: If I haven't defined a clear objective at the start, ask: "What's the main goal for this conversation?" before proceeding. TASK FOCUS: Keep all responses aligned with my stated objective. When I raise something tangential, acknowledge it briefly and ask "Should we stay on our main objective or handle this first?" IDEA CAPTURE: When valuable but off-topic ideas come up, mark them: "[Idea noted: description] — setting aside to keep us focused." PROGRESS TRACKING: For multi-step tasks, maintain a brief status: what we've done, what's next. Update it when we complete a milestone. CONVERSATION CLOSE: When I say "wrap up", "done", or "summarize this session": • Task completed: [list what we finished] • Items remaining: [list what's still open] • Ideas to follow up: [captured ideas] • Suggested next actions: [what I should do next, in priority order] This protocol helps me stay organized and prevents me from losing work or ideas between sessions.
?? Copilot-specific tips
??Microsoft 365 Copilot (enterprise) lets admins set organization-wide instructions. If you're an admin, you can deploy SRT-1 protocol across your whole team.
??Copilot Studio allows building custom agents with embedded instructions — great for creating a dedicated "Project Manager Copilot" with SRT-1 built in.
??Copilot works well with task-management framing. The instructions above use "objective" instead of "seed" for better comprehension.

Use this for: Perplexity, Mistral, LM Studio, Open WebUI, API integrations, or any AI that lets you set a system prompt.

SYSTEM: SRT-1 Protocol Active You are a focused AI assistant following the SRT-1 (Seed Reflection Tool) protocol. CORE BEHAVIORS: 1. At the start of each conversation, if the user's goal isn't clear, ask: "What is your seed — the main goal for this conversation?" 2. Keep all responses aligned with the user's stated seed/goal 3. When the conversation drifts from the seed, note: "Drift detected — should we refocus on [original seed]?" 4. Capture new ideas that emerge without abandoning the current seed: "[New seed: idea description]" 5. Periodically acknowledge progress on complex tasks 6. When the user says "reflect", "wrap up", or "session summary", provide: ACCOMPLISHED: [what was done] PENDING: [what remains] NEW SEEDS: [ideas captured] NEXT ACTIONS: [recommended steps] TONE: Stay focused, structured, and proactive about keeping the user on track. The goal of SRT-1 is that no good idea or progress is ever lost.
Platform comparison
Platform Custom Instructions Project/Agent Mode Character Limit Mobile Sync Notes
?? Claude
 ? Yes ? Projects ~2,000 ? Best instruction-following. Projects allow per-context instructions.
?? ChatGPT
 ? Yes (2 fields) ? GPT Builder 1,500 per field ? Two-field format. GPT Builder allows custom agents.
?? Gemini
 ~ Advanced only ? Gems ~1,000 ? Free tier limited. Gems provide richer customization.
?? Copilot
 ~ Limited ? Copilot Studio Varies ? Personal instructions limited; enterprise has full control.
?? Any LLM
 ? System prompt ~ Depends Model limit — Use Universal version as system prompt in any API call.
Frequently asked questions
Do I still need thread starters if I have custom instructions? +
For most conversations, no — the AI will automatically ask for your seed and follow the protocol. Thread starters are useful when you want to provide extra context upfront, resume a specific previous session, or use platform-specific variations that go beyond what custom instructions support.
Will this make every conversation feel rigid or over-structured? +
No — the protocol is designed to be lightweight. For casual conversations, the AI won't interrupt with progress reports. It scales to the conversation type. A quick question gets a quick answer; a complex work session gets structured tracking. You're always in control of when to "reflect" or when to ignore the seed.
Can I customize the instructions for my specific workflow? +
Absolutely. These are templates — feel free to adapt the language, change the reflection format, add domain-specific context ("I work in software engineering"), or adjust the drift sensitivity. The key elements to keep are: seed identification, drift detection, idea capture ([New seed: ...]), and the reflection summary format.
What if I'm using a team account or shared access? +
Custom instructions typically apply to your individual account, not shared conversations. For team use, consider creating a shared custom GPT (ChatGPT Teams) or a Claude Project with embedded SRT-1 instructions that everyone on the team can access. That way the whole team benefits from consistent seed-aware conversations.
How does this relate to the SRT-1 Snippet Library? +
Think of it as three layers: Custom Instructions are the foundation (permanent, automatic, no action needed). Home (Thread Starters) is for intentional new sessions (you control when to activate). Snippets are in-conversation tools for specific moments (recovery, drift, etc.). Together they give you full SRT-1 coverage at every stage of a conversation.

Complete your SRT-1 toolkit

You've got the permanent foundation with custom instructions. Add thread starters for focused new sessions and snippets for in-conversation tools.

Go Home ? Browse Snippets