Audits a codebase for coupling imbalances using Vlad Khononov's Balanced Coupling model and writes an architecture review document.
Guides Claude in creating and training self-improving agents using AgentDB's nine reinforcement learning algorithms.
Practical patterns for giving AI agents session memory, long-term storage, pattern learning and context management with AgentDB and ReasoningBank.
A skill for validating designed protein sequences and binder–target complexes with AlphaFold2/ColabFold and interpreting confidence metrics.
A runbook skill for deploying TestDriver test infrastructure to AWS via CloudFormation, with EC2 instances spawned and terminated automatically per Vitest test.
Teaches Claude AgentDB's advanced stack — QUIC sync, hybrid vector+metadata search, multi-DB sharding, MMR and production hardening.
A hard-won playbook for adversarially testing the fdeops CLI's engagement resolution, registry, hooks and <private> redaction boundary inside a throwaway sandbox.
A disciplined playbook for splitting work across multiple agents with dependency graphs, handoff contracts, approval gates, and verified synthesis.
An analyst-grade skill that decomposes how a public company makes money using XBRL segment data and SEC filings, with mandatory inline citations.
Builds a citation-backed earnings sentiment report by comparing analyst estimates against company guidance across recent quarters of SEC filings and earnings calls.
Mines SEC filings and earnings calls to decompose a company's organic and inorganic growth drivers into a citation-backed, analyst-grade report.
A skill that turns Wikipedia's editing tradition (NPOV, summary style, MoS) into concrete criteria for writing and reviewing neutral wiki reference pages.
A gated end-to-end runbook for releasing Compass: sync main, bump versions, open the PR, tag the merge commit, build six platforms, and verify every download.
An end-to-end npm release runbook: local verification, tag push, approval-gated GitHub Actions publish, then post-publish verification and cleanup.
An autonomous 13-step loop that pulls the next TaskMaster task, dispatches a complexity-tiered subagent, and only marks it done after evidence-gated triple verification and a reachability sweep.
Orchestrator skill that runs task waves across isolated git worktrees, verifies each worker branch by evidence, merges sequentially, and ends with a single PR.
A top-level orchestrator that turns a research direction or topic into a full package by chaining exploration, literature survey, experiments, and paper writing in order.
Takes a proven vulnerability through private vendor reporting, timeline coordination, CVE assignment, and public advisory publication.
A five-gate Socratic workflow that drives a one-line idea into an approved spec with single-test acceptance criteria before any code is written.
An on-demand audit that finds provenance-tracked docs gone stale from out-of-commit source changes and lets you re-scout, re-baseline, or defer each one.