suleymanbyzt
UserStay in control of AI-written code without drowning in implementation details. Works with Claude Code, Codex, and any AGENTS.md-compatible agent.
Categories
Indexed Skills (9)
code-review
Use when reviewing code you did not write in this task — a pull request, a teammate's branch, a diff, or another agent's output. Enforces risk-ordered reading, evidence-backed findings, and severity triage instead of a comment flood.
dependency-upgrade
Use when upgrading, adding, or removing a dependency, package, library, or framework version. Enforces reading release notes before bumping, mapping breaking changes to actual call sites, and proving the upgrade with a green suite — "it compiles" is not "it works".
performance-investigation
Use when investigating slowness, high resource usage, timeouts, or any optimization request. Enforces measure-first performance work — profile, prove the bottleneck, fix, measure again. No optimization based on guesses.
bug-fix
Use when investigating or fixing a bug, error, crash, regression, or unexpected behavior. Enforces reproduce-first, evidence-based root cause analysis, and a regression test before the fix.
feature-development
Use when implementing a new feature or extending existing functionality. Covers understanding the request, proposing a short plan, getting approval, implementing with tests, and recording the work.
integration-brief
Use after changing an API endpoint, contract, event, or message schema that another person or team will integrate with. Produces a short brief the integrating developer (frontend, mobile, another service) can implement from directly.
safe-refactor
Use when restructuring, cleaning up, or reorganizing existing code without changing its behavior. Enforces a test safety net before touching anything and proof that behavior was preserved.
verify-work
Use before claiming any task is done, complete, or fixed. A final verification pass that catches unverified claims, unintended changes, and leftover debris before reporting to the user.
record-decision
Use when a real architectural or structural decision is made — choosing between approaches, adopting or rejecting a technology, changing a contract, or any L3 task. Records the decision, the reasons, and the rejected alternatives in docs/decisions/.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.