Research workflow automation with Codex
Codex runs a weekly arXiv search, triages new papers, feeds a Zotero library, and turns hypotheses into cheap experiment tests for an AI safety researcher.
The author cut time on literature triage and experiment setup, and says related-work sections now need less citation double-checking. No numbers are given.
Overview
How it was set up
Codex app with weekly automations, a CSV of seen papers, Zotero with a vector database, subagent threads, Bitwarden Secrets Manager for scoped credentials, and a hook that blocks destructive commands.
Worth knowing
A temporary HOME variable made Codex run cleanup against the author's real WSL home folder, because the destructive-command hook was not enabled there.
A prompt to start from
Here is my hypothesis: [hypothesis]. Using the repo context and notes in [paths], propose cheap tests, confounders, success criteria and scrap criteria. Return paths, commands, logs and examples to inspect, plus the next cheapest check.
Written by Agentic Feats from the source. Not the author's exact prompt.
Sources 1
History
- Source published
- 2026-07-07
- Added to Agentic Feats
- 2026-10-13
- Pick
- No
- Markdown copy
- /c/research-workflow-automation-with-codex.md
More in research
36 featsResultIt created the text file as asked, but after clicking a result it took a screenshot too early and hallucinated the answer.
ResultOperator handled tour search and lesson plan but was slow; adding three grocery items took 15 minutes and an Uber price check took five.
ResultThe draft was usable but weaker than accepted NSF summaries; thinking mode helped, though it began hallucinating and misreading the sources.