nelsonwerd

Composable Agent Skills (Claude + OpenAI Codex) for taking an idea from fuzzy → validated → sequenced build → shipped — a manual tier (ideate, deep-dive, prompt-pack) and an autonomous tier (autopilot, build-loop, audit-and-fix).

Frozen artifact of a ~13.5h autonomous Claude run (idea-to-ship autopilot). Known unfixed SSRF, do not deploy. Case study, not a product.

Agent Skill for Claude and OpenAI Codex — take a fuzzy idea (or existing project) through explore → pressure-test → converge into a locked CONCEPT_BRIEF + roadmap. Upstream partner to prompt-pack.

Agent Skill for Claude and OpenAI Codex — break big, risky work into sequenced, self-contained prompts you run across separate chats, plus paste-ready handoffs.

Flight records for agent-written code — deterministic session capture, evidence ledger, and graded verification. No LLM in the trust path.

Run unmodified agent fleets (Claude Code, Codex, any subprocess) inside a corruption-evident, crash-durable event log — one local Go binary. kill -9 and resume, replay without re-executing effects, fork any timeline.

Rigorous multi-agent deep-dive analysis for Claude and OpenAI Codex — audits codebases, evaluates strategies/designs, tackles open research via parallel specialists → synthesis → adversarial red-team.