Explore Prompts

Design an agent that detects stale docs by comparing code/API schemas vs docs, then opens issues or PRs. Include a freshness score and SLA.

Create a release engineering agent: semantic versioning, changelog generation, tag signing, and release notes validation. Require human approval for major versions.

Design MCP wrappers that validate inputs, enforce types, and reject dangerous parameters. Include error taxonomy and audit logging for rejected calls.

Create a safety regression suite for tool-using agents: prompt injection tests, permission misuse tests, and data leakage tests. Gate deployments on this suite.

Design a dashboard that drives the agent’s priorities: complexity, test coverage, error hotspots, dependency risk, and SLOs. Include alert thresholds and weekly reports.

Create an agent that identifies unreliable dependencies and adds circuit breakers, timeouts, and fallbacks. Validate via chaos testing in staging.

Design a modularization plan: identify seams, extract modules, keep interfaces stable, and add integration tests. Require incremental steps and metrics.

Create a formatting agent that guarantees no semantic change: parse/print equivalence, tests, and diff classification. Include a strict policy for format-only PRs.

Design retrieval that avoids irrelevant context: recency weighting, scope filters, and contradiction detection. Include evals for context precision/recall.

Create review prompts that guide humans: “check invariants,” “check auth,” “check rollback.” Tune them based on escaped defects and reviewer feedback.

Design an agent to tune security headers and CSP safely: staged enforcement, report-only mode, breakage monitoring, and rollback. Require security team sign-off.

Create a plan to reduce logging costs: sampling, cardinality caps, and aggregation. Require proof that debugging capability is preserved via incident drills.