Mathematical and experimental audit workflows for SGFR, positive BWT, Class-IL metrics, replay geometry, and leakage checks.
Ships a Codex manifest but no marketplace file, so Codex can't add it by name yet. Add it to a local marketplace (.agents/plugins/marketplace.json).
Plugin sgfr-math-auditor in bytetiff/SolanumCL. Commands read from the repo's manifest — 1 month ago.
bwt-auditUse this skill when analyzing SGFR, Class-Incremental Learning metrics, positive BWT, forgetting, FAA, task accuracy matrices, or unusually strong backward transfer.
leakage-auditUse this skill when checking SGFR experiments for data leakage, split overlap, hash duplicates, label mapping errors, replay/test overlap, or anonymization issues.
protocol-auditUse this skill when checking whether an SGFR or Class-IL experiment follows the intended task order, no-task-id inference rule, checkpoint evaluation protocol, and metric averaging convention.
replay-geometry-auditUse this skill when analyzing SGFR feature geometry, SVD energy, replay exemplar selection, prototype drift, feature margins, class separability, or correlations between geometry and BWT.
theory-reviewUse this skill when reviewing or writing the mathematical explanation for SGFR, positive BWT, frozen-backbone replay, SVD selection, or task-balanced sampling.
General mathematical research audit plugin for formalization, proof checking, numerical validation, statistical testing, spectral geometry, concept disambiguation, cautious concept discovery, and adversarial research judging.
16 skills
Cross-session memory and token savings for coding agents.
6 skills · 1 agent · 1 MCP server · hooks
Nightly offline self-evolution for this repo's Claude agent: harvests past Claude Code sessions (read-only), mines recurring tasks, replays them offline on your own API budget, and consolidates learnings into validated CLAUDE.md memory and SKILL.md skills behind a held-out gate, staged for review (never auto-applied). Verbatim vendor of microsoft/SkillOpt's stdlib-only skillopt_sleep engine + Claude Code plugin surface (MIT).
1 skill · 1 command · hooks
Incrementally learns durable user preferences and workspace facts from transcript changes and keeps AGENTS.md up to date with plain bullet points.
1 skill · 1 agent · hooks
Nightly offline self-evolution for your local Claude agent: review past sessions, replay recurring tasks on your own budget, and consolidate validated CLAUDE.md memory + SKILL.md skills behind a held-out gate. Self-contained (vendored, zero third-party deps).
1 skill · 1 command · hooks
Use when a neural network grows, prunes, composes modules or adapts across tasks and needs lifecycle, gradient-isolation or state-integrity checks.
1 skill · 2 commands · 1 agent
This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.