Explore Prompts

Design explicit stop rules: max iterations, max diff size, max retries, and “ask human” triggers. Include monitoring for loop detection and runaway behavior.

Create a cost-aware policy: budgets per run, per tool call, and per environment. Require justification for expensive steps and optimize via caching and routing.

Design compatibility tests between LLM prompts and MCP tools: schema conformance, error handling, and anti-injection checks. Gate prompt/tool updates on these tests.

Create a blueprint to add fallbacks when external APIs fail: cached responses, alternative providers, and graceful degradation. Include policy for when to fail fast.

Design an agent that optimizes tail latency: identify bottlenecks, reduce contention, add caching, and validate with load tests. Track p95/p99 improvements.

Create a process to generate integration tests from anonymized traffic patterns (no PII). Require privacy filters, sampling, and periodic review.

Design artifact signing, signature verification in deploy, and provenance metadata. Ensure the self-edit loop cannot bypass signing checks.

Design a dead-code removal agent: identify unused code, verify no runtime references, remove incrementally, and monitor. Include rollback if hidden dependencies appear.

Create a plan to tighten interfaces with types, schemas, and validation at boundaries. Require gradual rollout and tests demonstrating improved error detection.

Design a service catalog that the agent maintains: owners, dependencies, SLOs, runbooks, and deploy pipelines. Use it to route reviews and risk analysis.

Create a system where the agent detects recurring insecure patterns and generates internal training snippets and secure alternatives, linked to real PR examples.

Design improvements to internal developer tools: faster local setup, better error messages, and templates. Use surveys + telemetry to prioritize changes.