paper2code

Use Codex Autoresearch when a repository task has a numeric outcome, such as fewer test failures, better coverage, lower latency, or a smaller binary. The skill confirms the target, editable scope, metric command, regression guard, and run mode, then lets Codex try one focused change per iteration. Improvements are committed and retained; failed or regressive trials are reverted through Git, with an auditable event history. It can run interactively or in the background, but each experiment requires a clean named Git branch and sufficient access to create and revert commits.

Agent Sandbox Skill gives Claude Code, Codex, and Gemini CLI managed E2B environments for engineering work that should not touch the local filesystem or production systems. Agents can install dependencies, run untrusted code, scaffold and build full-stack applications, expose a hosted port, and perform Playwright-based visual checks. Commands cover small sandbox tasks as well as plan, build, host, and test workflows, with lifecycle tools that preserve a sandbox across turns. An E2B API key is required, and compute and hosted previews depend on the E2B service.

Goal Forge interviews a rough idea into a SPEC.md with user-approved, measurable completion criteria, challenges ambiguous requirements, compiles the result into a structured GOAL.md, and performs a read-only Codex configuration check. Its contract captures a scorecard, fast feedback loop, slower final validation, working-memory files, and an optional control surface so multi-hour /goal sessions can remain coherent. The suggested unattended configuration can grant broad filesystem access and disable approvals, so it is intended only for project paths the user has explicitly trusted.

RecallLoom adds a file-based memory layer to long-running AI work, storing project background, current progress, key decisions, and next steps in Markdown and JSON beside the project. When users switch sessions, models, or coding tools, a new agent can restore the controlled current state before continuing, record progress, and check whether continuity files are missing, stale, or conflicting. It is deliberately focused on project continuity, not repository-wide knowledge retrieval, task orchestration, or autonomous execution.