Public-Portfolio-Challenge / lockbox-holdout
AGENTS.md82 starsChanged 3 months agoMITView on GitHub
The single-touch lockbox — a final anti-overfitting holdout run once, after design freeze, on a window held out from every fold, sweep, and search. Use when finalizing a bakeoff winner before deploy, setting up the A/B/C baselines as OOS bars, or running the S1.5 gate-coherence auto-relax. Covers why looking at the lockbox twice burns it, the lockbox pass conditions, and the three baselines. The lockbox is distinct from the walk-forward OOS folds. Invoke with /lockbox-holdout, or let it auto-activate when…
In this repository23 other files it gives its agents.
- skills/alt-data-indicators/AGENTS.mdAGENTS.md
- skills/breadth-audit/AGENTS.mdAGENTS.md
- skills/bug-protocol/AGENTS.mdAGENTS.md
- skills/deploy-gate/AGENTS.mdAGENTS.md
- skills/engine-sanity/AGENTS.mdAGENTS.md
- skills/options-structure-rules/AGENTS.mdAGENTS.md
- skills/portfolio-certification/AGENTS.mdAGENTS.md
- skills/run-episode/AGENTS.mdAGENTS.md
- skills/strategy-bakeoff/AGENTS.mdAGENTS.md
- skills/sweep-reoptimization/AGENTS.mdAGENTS.md
- skills/walk-forward-oos/AGENTS.mdAGENTS.md
- alt-data-indicatorsSkill
- breadth-auditSkill
- bug-protocolSkill
- deploy-gateSkill
- engine-sanitySkill
- lockbox-holdoutSkill
- options-structure-rulesSkill
- portfolio-certificationSkill
- run-episodeSkill
- strategy-bakeoffSkill
- sweep-reoptimizationSkill
- walk-forward-oosSkill
The file
Files
lockbox-holdout
The single-touch lockbox — a final anti-overfitting holdout run once, after design freeze, on a window held out from every fold, sweep, and search. Use when finalizing a bakeoff winner before deploy, setting up the A/B/C baselines as OOS bars, or running the S1.5 gate-coherence auto-relax. Covers why looking at the lockbox twice burns it, the lockbox pass conditions, and the three baselines. The lockbox is distinct from the walk-forward OOS folds.
Activation
Invoke with /lockbox-holdout, or let it auto-activate when a task matches the description above.
This is a SKILL.md skill on the open Agent Skills standard — Claude Code, Codex CLI,
Cursor, Gemini CLI, GitHub Copilot and ~15 other tools read it (natively or via the
install.sh adapter).
Usage
See SKILL.md for the full workflow, triggers, hard constraints, and detail. Reference material (if any) lives in references/.
Requirements
The NexusTrade MCP server must be connected — this skill drives NexusTrade's deterministic backtest/certification/compute tools (deferred; load schemas via ToolSearch).
Part of
The Public Portfolio Challenge skills library — a functional decomposition of the Episode 10 certification runbooks. See the library index at ../README.md.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.