Menu
Monitor local background experiment metadata, process state, logs, and result files.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Launch bounded runner experiments as local background jobs without SSH.
Reader role helper for turning shared runtime evidence into the next bounded objective.
Use when the user gives a vague research direction for a new sub-PHD run and wants one-question-at-a-time refinement that is correct but not overly precise, leading to a candidate rewrite of person_program.md before any final write.
Use when the user wants an evidence-first, staged briefing of the current sub-PHD run window, including what the run did, progress versus person_program.md, which propositions have meaningful conclusions, what evidence supports them, and plain-language explanations of new concepts.
Runner role helper for implementing and validating bounded changes in the repo working tree.
| name | local-watch |
| description | Monitor local background experiment metadata, process state, logs, and result files. |
Use this skill when research_agent.toml selects [execution].backend = "local" and the loop is monitoring a local run.
scripts/research_supervisor.py watch-backend --backend localwatch_status = "synced" as local evidence ready for reader or runner review; no remote sync is neededlocal_evidence_paths pointing only to local files