backfill
Skill2 starsChanged 8 days agoMITView on GitHub
Process past Claude Code sessions with Ouroboros — measure them (baseline of corrections) or learn from them as if Ouroboros had been active (proposals go to the approval queue).
5 other projects use a version of this.
Adopted by
Most recently taken up by
- clins1994/jobcan-agenttheir own version, first seen Nov 8, 2026
- ronniepinnell/caspertheir own version, first seen Nov 8, 2026
- scaccogatto/okf-skillstheir own version, first seen Oct 13, 2026
- chrisselig/claude-code-toolkittheir own version, first seen Oct 9, 2026
- iwe-org/marketing-workspacetheir own version, first seen Oct 8, 2026
Other versions5 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 5
- scaccogatto/okf-skills★ 444same name
- iwe-org/marketing-workspace★ 6same name
- chrisselig/claude-code-toolkit★ 1same name
- clins1994/jobcan-agent★ 1same name
- ronniepinnell/casper★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
Files
Backfill past sessions
Arguments: $ARGUMENTS — first the mode (measure, the default, or learn), then days to look back (default 30), then the maximum number of sessions (default 20).
- measure: counts corrections and lists what could have been learned; writes nothing. Skips sessions already recorded.
- learn: reviews each session as if Ouroboros had been active, oldest first, and sends every proposal to the approval queue (similar proposals from different sessions are merged, the newest version wins). Corrections are not counted again. Skips sessions already learned from.
Dry run first (add
--learnfor learn mode):node --no-warnings "${CLAUDE_PLUGIN_ROOT}/scripts/review-worker.mjs" --backfill --dry-run --days <days> --limit <limit> --data "${CLAUDE_PLUGIN_DATA}"- Tell the user how many sessions and user turns that is, and that each session costs one isolated small-model call (two in measure mode: the correction count and what would have been learned; typically under US$ 0.01 each with Haiku, plus small merge calls in learn mode; capped by the plugin's monthly budget). Ask for confirmation.
- On confirmation, run the same command without
--dry-run(run it in the background if it is long) and report the final lines (calls, cost and, in learn mode, how many proposals are waiting). In learn mode, point the user to the inbox:/ouroboros:inbox, or the Inbox page of/ouroboros:dashboard.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

