Step through versioning, tagging, and verification

A composable skill library that turns Claude Code or Codex into an orchestrator with parallel subagents, verification, and self-improving workflows.

Reduce negative-constraint and session-history leakage when a discarded proposal or user correction is echoed into final artifacts as a ‘without X’ label, rejected-option explanation, or process residue. Use for 此地无银三百两式 output in prose, code, metadata, and handoffs, including later requests to finish, commit, publish, or open a PR after iterative work; not for ordinary deletion, deprecation, migration, or requirements where the exclusion itself is material.

Steward the vanillagreencom/kendex issue queue (Linear team KEN is the poll surface; GitHub stays the PR/code surface) on a self-paced loop: watch open PRs, poll, triage (dedupe; close non-kendex issues and repost project-local ones to their owning repo; fix genuine defects in kendex's skills/agents/hooks/pi-extensions or the Rust CLI), run each fix through the orch skill, merge, propagate via kendex refresh, then reschedule. A thin wrapper: fix cycles belong to orch, PR mechanics to the github skill, PR monitoring to review-gate's pr-watch, Linear ops to the linear skill — this skill carries only the kendex-specific stewardship knowledge. Use when asked to monitor kendex's issues continuously or to run one fix-and-propagate cycle.

Diagnose and fix slow pyfixest regression GRIDS (many feols/fepois calls run sequentially) that stay slow despite demeaner_backend="cupy64" and an idle GPU. Use when: (1) a script looping dozens of pf.feols models on a 100k+ row panel takes ~1 min/model, (2) process inspection shows ~1-1.5 cores busy and nvidia-smi shows ~0% GPU utilization with a resident cupy context, (3) planning any worker prompt that will run a model grid (robustness variants x FE structures x domains). Root cause: per-model CPU-side single-threaded fixed costs (formulaic model-matrix build, interaction construction, singleton detection, cluster vcov) dominate wall time; GPU demeaning is a small slice. Fix: shard the model grid across OS processes and/or use pyfixest multiple-estimation syntax; mandate this IN THE WORKER PROMPT.

Build, edit, redesign, render, and verify polished editable PowerPoint `.pptx` decks from a prompt, structured JSON, local data, or a saved workspace. Use for scientific, lab, clinical, consulting, board, investor, editorial, policy, and operational presentations where narrative, visual hierarchy, readability, and reproducibility matter.

Scraping + AI-provider standards — the Scraper trait & SCRAPERS registry, selector resilience, rate-limiting/cancellation, and the provider-abstraction (zero-change) rule for embeddings/streaming/prompts. Load for changes under scraping/, ai_provider/, packages/prompts, documents/embed.

Resolve rebase and merge conflicts for an already-open Vigilante pull request, keep the branch validated, and report progress back to GitHub.

>-

Orchestrate parallel implementation and fresh validation agents safely in the Sero repository. Use when a task asks for pi-subagents, parallel workers, isolated worktrees, review-and-fix loops, or several independent fixes that must be integrated and reviewed without Claude Code.

Select, probe, calibrate, and audit the Burger viscoelastic contact model for PFC 5.0 asphalt-mixture cases, including Maxwell-Kelvin parameter roles, units, temperature/rate dependence, PFC5 cmat assignment, timestep checks, creep/recovery validation, and PFC5-only syntax boundary checks.