effortmining

Plugin effortmining from marketplace effortmining in nagisanzenin/effortmining. Commands read from the repo's manifest — 1 month ago.

Skills (2)

• effort-bench

  Drive the effortmining A/B benchmark harness, the deterministic Python core (bench/effort.py) that measures pass-rate and token cost per (task-class, effort-tier) and fits the calibration table the effortmine skill dispatches from. Use to validate the instrument, run the matrix, grade, analyze, repo

• effortmine

  Decompose a request into subtasks, classify each by difficulty (T1 mechanical through T4 hard reasoning, plus R-research and C-coding), and dispatch each to the cheapest calibrated reasoning-effort tier a blind grader still accepts. Use when you want a multi-part job done at the right effort per par

Subagents (6)

• effort-grader

  Blind grader for effortmining. Grades one artifact against a task prompt and a fixed rubric ONLY. The input payload deliberately carries no field naming the tier, agent, or model that produced the artifact, so the grade cannot be biased by effort level. Skeptic-first, rounds scores down, emits stric

• miner-high

  Delegate worker pinned to HIGH reasoning effort, the Opus 4.8 model default. Pick for diagnosis, multi-constraint logic, or tracing where the work is figuring out what and the answer is short. Pilot v1 (2026-07-06) calibrates this tier for T3-moderate-reasoning — the class with the real quality grad

• miner-low

  Delegate worker pinned to LOW reasoning effort, the cheapest tier. Pick for mechanical retrieval, extraction, and reformatting with an explicit output template and no reasoning required. Fitted table (v1+v2, 2026-07-07) calibrates this tier for T1-mechanical, T2-simple-transform, and — with fit-blin

• miner-max

  Delegate worker pinned to MAX reasoning effort, the most expensive tier and session-only. Pilot v1 (2026-07-06) calibrates no class to this tier and found saturation at max in all four classes — it never improved quality over xhigh and only cost more (no strict regression observed). Reserve it for w

• miner-medium

  Delegate worker pinned to MEDIUM reasoning effort. Calibrated for C-coding since the 2026-07-07 refit (the C6 invariant task breaks low; medium sits at the quality ceiling at ~1/7th of max's tokens).

• miner-xhigh

  Delegate worker pinned to XHIGH reasoning effort. Pick for hard, multi-step reasoning with adversarial or overfit-punishing edge cases such as combinatorial counting, constrained ordering, or ledger-style state. Pilot v1 (2026-07-06) calibrates no class to this tier — its T4-hard-reasoning tasks wer

Hooks (2)

• SessionStart

  ${CLAUDE_PLUGIN_ROOT}/hooks/session-start.sh

• PostToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/log-dispatch.sh

Similar plugins