Pieces Prompts

Act as my developer-productivity coach. I keep losing useful code across [IDE], browser tabs, and chat threads. Show me how to use Pieces long-term memory to fix this: first, install PiecesOS and enable Long-Term Memory in the tray menu; second, work normally for [NUMBER] days so it captures snippets, error messages, and docs in context; third, use the Pieces copilot to ask 'show me the [LANGUAGE] function I saved around [DATE] that handled [TASK]' and convert the best results into a named, tagged snippet collection for [PROJECT]. End with a weekly routine (under [NUMBER] minutes) that keeps the library curated instead of cluttered.

Act as an engineering manager. A new [ROLE] joins our [TEAM SIZE]-person team on [DATE] and must become productive in [PROJECT] within [NUMBER] weeks. Design an onboarding plan built on Pieces: the new hire installs PiecesOS with long-term memory on, pairs with a mentor whose saved snippets and error fixes are shareable, and uses the Pieces copilot to ask 'how did we solve [RECURRING ERROR] last time in [REPO]?' instead of pinging Slack. Include three specific memory queries the new hire should run in week one, and a checklist for the mentor to tag their most valuable captured context for reuse.

Act as a senior [LANGUAGE] developer. Every few weeks I hit [ERROR TYPE] in [PROJECT] and re-derive the same fix. Build me a Pieces workflow: configure PiecesOS long-term memory to capture error messages and terminal output automatically while I work in [IDE]; when the error recurs, ask the Pieces copilot 'what did I try last time [ERROR TYPE] appeared, and which fix actually worked?' Then show me how to enrich the saved snippet with a note explaining the root cause in [NUMBER] sentences, tag it [TAG], and share it with [TEAMMATE]. Finish with how to use Pieces search filters (keyword, date, project) to pull the fix in under 30 seconds.

Act as a staff engineer. I maintain [NUMBER] repositories and keep rewriting the same [PATTERN TYPE] logic (for example [EXAMPLE]). Show me how to use Pieces to stop: let long-term memory capture the implementations across [REPO A] and [REPO B]; then use the copilot to compare them and identify the most robust version; then extract it into a shared snippet with placeholders for [VARIABLE 1] and [VARIABLE 2]. Include the exact questions to ask Pieces to surface all variants, and a short rubric for deciding which variant becomes the canonical one our team reuses.

Act as a developer evaluating AI models. I want to know whether [MODEL A] or [MODEL B] explains [LANGUAGE] code better for my use case. Give me a repeatable test using Pieces multi-LLM switching: pick [NUMBER] saved snippets from my Pieces memory covering [TOPIC 1], [TOPIC 2], and [TOPIC 3]; ask each model the same question ('explain this code and flag the riskiest line') through the Pieces copilot; score the answers on accuracy, clarity, and actionability in a simple table. Tell me how to keep the test snippets private (on-device processing) and how often to re-run the comparison as models update.

Act as a technical writer. Our team has [NUMBER] months of captured workflow context in Pieces but our docs for [PROJECT] are stale. Show me how to convert it: use the Pieces copilot to pull the most-asked questions and most-saved snippets about [MODULE]; draft a [PAGE COUNT]-page runbook with sections for setup, common errors ([ERROR 1], [ERROR 2]), and the [WORKFLOW] workflow; then have the copilot verify each step against the actual captured context so nothing is invented. End with a monthly refresh routine that re-pulls new memory items and flags doc sections that went stale.

Act as a senior reviewer. I have to review a [NUMBER]-file PR in [REPO] touching [MODULE] and I did not write the original code. Give me a Pieces-powered prep routine: ask the copilot to recall my saved context about [MODULE] from the last [NUMBER] months (design decisions, past bugs, related snippets); generate five targeted review questions about the riskiest changes; and draft review comments that reference the actual history ('we fixed [BUG] here in [MONTH]; this change reintroduces the pattern'). Keep the whole prep under [NUMBER] minutes and flag anything the memory suggests is a repeat of a past incident.

Act as a mentor for a [LEVEL] developer learning [TECHNOLOGY]. I want to use Pieces as a learning journal over the next [NUMBER] months. Design the system: tag every saved snippet with what I learned ([CONCEPT]) and a confidence rating from 1 to 5; each Friday, ask the Pieces copilot 'summarize what I struggled with this week in [TECHNOLOGY] based on my saved errors and searches'; each month, have it quiz me on the lowest-confidence items with fresh practice problems. Include the exact Friday prompt template and a rule for when a topic graduates from the journal (confidence [NUMBER]+ on three separate weeks).

Act as an AI engineer. I run [AGENT TOOL] and want it to read my Pieces long-term memory through the MCP protocol. Walk me through it: confirm PiecesOS is running with long-term memory enabled; configure the Pieces MCP server as a tool source in [AGENT TOOL]; then test with a query like 'what do I know about [TOPIC] from my past [NUMBER] months of work?' Show me how to scope what the agent can see (exclude [SENSITIVE PROJECT] via opt-out controls), and give me three agent workflows that get dramatically better with memory access (for example: 'draft the [DOCUMENT] using my past decisions on [TOPIC]').

Act as a developer who lives in meetings. My [MEETING TYPE] meetings produce decisions about [PROJECT] that I forget by the time I code. Build a Pieces workflow: capture the meeting notes and whiteboard snippets as memory items tagged [PROJECT] and [DATE]; when I sit down to implement [FEATURE], ask the copilot 'what did we decide about [FEATURE] in the [DATE] meeting, and what constraints did [STAKEHOLDER] mention?'; then generate the implementation scaffold with those constraints baked in. End with a template for the capture note so every meeting produces a memory item the copilot can actually retrieve later.