hubermjonathan/agents
Run agents and improve their skills: parallel workers, decision logs, session transcripts, and skill audits
Measure from past session transcripts how often each skill loads when the work calls for it, then propose fixes for the ones that miss.
Find this session's transcript file, or list where agents keep session transcripts. Use when a skill or task needs the full record of the current session, or needs to read past sessions.
Hold off on changes to a topic while you look into it, until the user settles it.
Mine a session, this one by default, for durable learnings and turn each into an edit to a skill, an agent instruction file, a setting, or a new skill. Works mid-session.
Keep a reviewable decision trail for long-running or unattended work: a TSV log with one row per decision (what, why, evidence, result). Use for autonomous or multi-phase runs, work a human reviews after stepping away, or when another skill needs an audit trail.
Fan out N parallel workers, drain them, and return one report. Use for multi- reviewer code review, parallel coverage of separate slices, races on the same brief, or parallel exploration, and when another skill needs parallel workers.
Writing documents and prompts for agents. Use when creating or editing skills, modifying AGENTS.md or CLAUDE.md, or writing a prompt for a subagent, a workflow agent, or another session (a handoff prompt).