Write state-restoration documents for passing tasks between agents or engineers.
Use when the task is to understand an unfamiliar codebase, locate key entry points, or summarize architecture before editing.
Check existing repo capability, external libraries, MCP options, and maintenance risk before writing custom code. Decide adopt/wrap/build with explicit criteria.
Standardize the reporting of model metrics to ensure statistical rigor and business relevance.
Audit whether a dbt incremental model uses the right incremental strategy for the repo, the data shape, and the operational constraints.
Use continuously during long tasks. Teaches how to read less, output less, and keep the LLM context window lean and fast.
Use when diagnosing agent session history, interrupted tool loops, missing tool results, timing bottlenecks, or subagent trace correlation.
Audit agent configuration files for security vulnerabilities and misconfigurations. Covers settings.json, .mcp.json, .codex/config.toml, AGENTS.md, hooks, plugin manifests, and…
Use when an MCP server is about to be enabled, added to agent config, or upgraded -- a staged server config is a candidate for activation, a third-party server is proposed from a…
Analyze BigQuery usage, identify cost hotspots, repeated failures, and practical optimization opportunities.
Diagnose local agent and project setup problems — missing commands, broken MCP configs, stale generated assets, bad paths, invalid catalog entries, and missing env vars.
Diagnose NaN losses, out-of-memory errors, and shape mismatches in deep learning or ML pipelines.
Full autonomous execution from idea to working code — from yeaight7/agent-powerups
Use when about to commit, push, or publish -- staged changes touch config or environment files, generated artifacts (relay sessions, logs, build output) are being added, or the…
Relay questions and tasks to a persistent local Gemini ACP session with cross-turn context.
Run large codebase migrations in reviewable local batches with codemods, checkpoints, and verification. Use when a wide refactor is too risky to ship as one monolithic change.
Relay questions and tasks to a local Claude CLI using one-shot subprocesses per ask.
Ensure that new complex models actually outperform simple, naive baselines.
Process-first advisor routing with artifact capture
Extract a learned skill from the current conversation manually
Turn vague implementation requests into testable requirements before coding.
Multi-model orchestration — route to two external advisors, then synthesize
Use to diagnose tests that pass and fail intermittently without code changes.
Identify and prevent target leakage in ML preprocessing pipelines.
Verify that API endpoints match their OpenAPI/Swagger specifications.
Use when code needs restructuring without changing observable behavior.
Audit the methodology used to split data into train, validation, and test sets.