Explore Prompts

Design a routing agent that selects specialized agents (tests, security, perf, docs) based on task intent and risk. Include confidence thresholds and fallbacks.

Create a linter for agent prompts: ban risky instructions, require citations for factual claims, enforce “no tool calls from untrusted text” policies.

Design a data quality gate: schema, drift detection, missingness, and label quality checks. Integrate with self-improving code changes to ML pipelines safely.

Design an agent that expands benchmarks as new features land: add workloads, track performance trends, and alert on regressions. Include benchmark governance.

Create a plan for generating API clients: typed SDKs, version pinning, integration tests, and docs. Include a policy against generating unsafe dynamic code.

Design an agent that writes release notes from diffs and PR descriptions, focusing on user-visible impact, migrations, and known issues. Include accuracy checks.

Create a blueprint for self-editing CI/CD scripts safely: validate configs, run in sandbox, and require manual approval for permission changes.

Design an ADR generator that proposes ADRs for significant changes: context, decision, alternatives, consequences. Require human approval and link tests/metrics.

Create a doc search agent that indexes internal docs, code comments, and ADRs, then answers “why is this here?” with citations for maintainers.

Design an incremental modernization strategy (language/runtime upgrades) with compatibility tests, staged rollouts, and metrics. Include “stop if risk” criteria.

Create an agent that turns bug reports into minimal reproducible cases, adds regression tests, and proposes fixes. Include a triage rubric and severity mapping.

Design a process where the agent proposes new lint rules only with evidence of repeated issues. Include rollout plan, false-positive checks, and developer ergonomics.