Correction Pipeline
Traces a user-flagged factual error to its source and fixes it in the brain, memory, or facts table.
- Repository
- garrytan/gbrain
- Stars
- 31,447
- Licence
- MIT
- Files
- 1 file
What it does
When you tell the agent it got a fact wrong, this skill makes it stop, trace where the mistake came from, and fix the original source instead of just noting it. It checks the gbrain knowledge store, the agent's memory files, its identity files (SOUL.md and USER.md), and the facts table, then classifies the error (bad data, stale data, cross-attribution or hallucination) and edits the right place. It also tries to write a guard note so the same mistake is less likely to recur.
Great for
- Developers running the gbrain personal knowledge store who want wrong facts corrected at the source
- Teams maintaining an AI agent's memory and people/company pages who need an audit trail of corrections
- Users who want the agent to explain why it made an error, not just apologise
Skip it if
- Simple one-off chats with no persistent brain or memory files, since most steps assume gbrain is installed (the skill is inert or confusing without it)
- Anyone who wants the agent to edit SOUL.md or USER.md by hand: the skill forbids that because hand edits are overwritten on the next render
- Sensitive records where automatic edits and commits to a knowledge repo are unwanted without review, because the skill writes files, commits, syncs and expires facts
Try asking
- “That's wrong, Sarah Chen joined Acme in 2021, not 2019. Find where that came from and fix it.”
- “You got that wrong. Where did you get that? Root-cause it.”
- “I never said that about my budget. Trace the error and prevent it happening again.”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 1 original files, the MIT licence and an attribution note.
correction-pipeline.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip correction-pipeline.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip correction-pipeline.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Name the wrong claim precisely so the search steps find the right rows and pages quickly.
- Review the commits and facts-table changes the agent makes, since it edits and expires data on its own.
- For identity-file errors, let the agent route through the soul-audit skill rather than editing rendered files directly.
What's inside
- SKILL.md
Redistributed unmodified from garrytan/gbrain under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.