Measures runtime, memory, and cold-start benchmarks and detects regressions against recorded baselines.
npx skills add https://github.com/jokerman89/lintel --skill perfbench-jokerman89
Please help me install this Agent Skill. Skill: perfbench Source: https://github.com/jokerman89/lintel/tree/main/skills/perfbench Command: npx skills add https://github.com/jokerman89/lintel --skill perfbench-jokerman89
High-intent search queries and answers about installing and using this skill.
Run the benchmark suite against the last recorded baseline; each scenario's median, p95, and memory are diffed automatically. Regressions are flagged at +5% (warn), +15% (P2), and +30% (P1) so you can address severe cases before release.
Pass the scenario name with --scenario and increase sampling with --iterations, for example 20 iterations for tighter percentiles. Warmup iterations run first and are discarded so measurements reflect steady-state behavior.
It detects a bench script in package.json, pytest-benchmark or hyperfine via pyproject.toml, or a repo-specific bench/ directory. If none exists, you can declare a suite in ~/.lintel/perfbench.yaml.
Use a profiler when you need to find the root cause of a bottleneck, since benchmarking only measures and compares results. The intended flow is to flag a regression with benchmarks, then hand off to a profiling subagent for investigation.
Single-iteration runs produce high noise; use at least 3 iterations with warmup runs discarded. Comparing results across different machines, such as a laptop versus CI, also produces misleading diffs.
optaveCompare current benchmark results against a saved baseline to detect performance regressions.
jerrywu001Investigate software errors through a four-phase root-cause workflow.
TravisLeeeeeeAutomate API endpoint validation for correctness, security, and performance.
zeuikliProfile applications to identify dominant bottlenecks and produce before/after performance reports.
