measurement plugins

by ML.ENERGY

Measure the energy and power consumption of code with Zeus, on NVIDIA and AMD GPUs, Intel/AMD CPUs and DRAM (RAPL), Apple silicon, and NVIDIA Jetson.

by thousandflowers

Measures what share of your agent's loaded context is ever actually invoked - skills, MCP servers, subagents, hooks, always-loaded prose - using evidence from your own session transcripts. Then prunes the dead weight, reversibly.

by Melodic Software

Read-only code measures for a change, with cited references and no verdict: lines per file (audit-size), cyclomatic, cognitive, and Halstead complexity (audit-complexity), duplication (audit-duplication), per-function coverage and CRAP from existing lcov, Cobertura, coverage.py, or Go artifacts (audit-coverage), TypeScript and Python type debt (audit-type-debt), metric literacy (principles), and setup. Uses only collectors already installed; never installs or runs tests.

by Melodic Software

Measures a Claude Code session's fixed startup context payload per item at a pinned binary version, including per-tool attribution of the built-in tool pools that /context reports only as totals, derived by A/B deny differencing under comparability rules. Falls back from an SDK meter to a headless /context parser, then to an error rather than a wrong number. Keeps a per-project before/after ledger per lever. Report-only by default; fix applies one approved project-scope trim.

by Melodic Software

Measurement-first optimization that refuses to report what the data does not support. go-faster sweeps the whole process for evidenced ways to go faster. target ranks candidates by evidence quality. goal sets realistic and ideal targets and computes the floor first. snapshot captures interleaved baseline and post runs on a qualified host and refuses wall-clock claims from a noisy one. verify re-derives results in a fresh context. protect locks in a counter win with a CI ceiling.

by Melodic Software

Outcome-verification stage: prove a change achieved its intended outcome (`/verification:confirm`: a mechanical build/test/lint prerequisite gate, then intent-match + evidence + verdict with the criterion auto-detected by change type), and verify measurable-improvement claims against a planning-time baseline (`/verification:measure`), never fabricating numbers.

by Grimaldo Stanzani

Discipline for evaluation acts: structure an experiment and its write-up so the Methods reconstruct without the conversation and the uncertainty is declared, not hidden — a typed record.yaml across a probe / measurement / decision tier ladder, with every load-bearing rule a script that exits non-zero rather than a line of prose.

by Vivswan Shah

Controls before a zero or absent reading, an alarming finding, a success claim, or stillness becomes a conclusion: evidence-bearing probes, positive and negative controls, second derivations, postconditions over logs, and a re-measure at the moment of sending.

by Webgrip

Refactor agent-maintained code where it measurably lowers what coding agents spend on future changes: git-hotspot targeting with agent share, a repeated-run token measurement harness over the agent's own usage stream, Fowler-catalog refactorings ordered by what shrinks the read set, tool-driven execution with a per-step gate, and a break-even model priced per token class with a safety factor.