Prompts that make your agent sharper
Battle-tested prompts for AI coding agents. Official picks are curated by the maintainers; community submissions are reviewed before publishing — so every prompt here is one someone actually shipped with.
Browse
Filter by agent, category, and what's trending or official.
Fill the blanks
Prompts expose {{variables}} — fill them in, no editing required.
Copy
One click copies the finished prompt to your clipboard.
Run
Paste into Claude Code, Codex, or any agent and go.
Security review a diff before merge
A checklist-driven security pass on a diff — verdict per item with file:line evidence, ending in a block/allow call. No vague summaries.
Flaky CI failure: classify, then one experiment
Turns an intermittent CI failure into a diagnosis: classification with log evidence, ONE confirming experiment, and only then a fix candidate.
Answer structural questions before touching code
Forces the agent to map callers, blast radius, and contracts before editing an unfamiliar symbol — with whatever code-intelligence it has.
No code edit without a traced spec
Before writing code, the agent must anchor the change to a spec/ADR/issue — or write the missing doc first. Intent must survive the author.
Adversarial self-review
Review your own diff as a hostile senior reviewer before handing it over.
Plan before code
Force the agent to separate problem, behavior, rules, and risk before any edit.
Context budget guard
Keep the agent from drowning itself — read only what the task needs.
Verify before you claim done
No "done" without an executed verification command and its real output.
Session handoff digest
Persist what the next session needs — not what this session did.
Minimal repro before any fix
Refuse to fix until the bug is reproduced in the smallest possible harness.