Skip to complete skill
Free community skill@ataiiam

User-correction learning loop

Independently compiled by XSkills from public material by Atai Barkai.

Pair an agent's miss with the user's corrective action, distill the pair into readable procedural memory, review it separately, and make it available to the next matching run.

Skill brief

Is this the right skill?

Use it when

  • You operate an agent in a domain where automatic right-or-wrong scoring is incomplete.
  • You can capture both the agent action and an authorized user's correction in the same workflow.
  • You can review, version, narrow, and remove learned procedural context.

What you get

  • Paired agent-miss and user-correction trace
  • Proposed procedural-memory lesson
  • Test and human approval state

Workflow

7 moves
  1. 01

    Choose the learning layer

  2. 02

    Capture the paired trace

  3. 03

    Distill procedural context

  4. 04

    Keep the acting agent separate

  5. 05

    Review before the lesson lands

  6. 06

    Use the readable lesson later

  7. 07

    XSkills adaptation: govern the memory

Complete instruction set

Complete SKILL.md

Read every instruction below or copy the complete file directly into your agent.

User-correction learning loop

Source workflow

Choose the layer you own

Separate three learning layers:

  • model: trained weights;
  • harness: loops, tools, prompts, and checks around the model;
  • context: memory and skills stored outside the model.

In product workflows without a free, trustworthy automatic score, work in the harness or context rather than the model. Context is readable plain text that can be changed, deleted, shared, moved to another model, or rolled back.

Capture both halves of the correction

Capture the agent and user in the same interface. Keep the agent's proposed decision together with the user's corrective action. A screen recording sees the person's actions but not the agent; a chat correction sees the words but misses the person's other actions.

Use the interaction event stream to retain the tool calls, state changes, approvals, the agent's miss, and the person's fix in one trace. The source uses AG-UI and CopilotKit as the implementation example.

Distill the pair into procedural context

Have a background agent record what the person did and why. Write it as procedural memory-the way to handle a matching case-rather than changing model weights.

In the source's refund example, the agent refuses a $5,000 refund because of a $2,000 limit; a manager approves it for a loyal customer after a third late delivery. The paired state, decision, action, and reason form the lesson.

Review before the lesson lands

Keep the acting agent from editing its own core memory. Use a separate agent or maintenance process to propose the context change. Preserve the source's Hermes boundary: a proposed skill change passes tests and a person before it lands.

Read the lesson on the next matching run

Let the next agent read the approved procedural memory when a case looks the same. Keep the text readable and reversible so a reviewer can inspect, delete, or roll it back.

XSkills operational adaptations

Preflight access to the interaction-event stream, a writable procedural-memory store, a separate proposal process, tests or replay cases, and a human approval step. If any capability is unavailable, return the paired trace and proposed lesson without claiming it was installed, replayed, or monitored.

Record a narrow trigger, action, exceptions, evidence, and scope for the proposed lesson. Recheck stale lessons and counterexamples before reuse. These schema and lifecycle mechanics are XSkills additions, not Atai's stated procedure.

XSkills safety adaptations

Obtain an appropriate consent and access basis before capturing user interactions. Minimize unrelated messages, personal data, and secrets; define retention and deletion. Do not learn illegal, discriminatory, protected-trait, or policy-violating instructions. Require qualified human review for high-impact decisions and explicit approval before a context change is published to users or teams.

Return

Return the minimized paired trace, proposed procedural lesson, capability status, test or replay evidence when available, human approval state, and installed version or honest proposal-only fallback.

Source attribution

Independently compiled by XSkills from Atai Barkai's X Article explaining self-learning for agents. Barkai did not author or review this skill. Capability fallback, lesson schema, replay and staleness checks, consent, minimization, retention, prohibited-learning limits, high-impact review, and publish authority are XSkills adaptations. Attribution does not imply endorsement.

Keep exploring

Related agent skills

Browse ai workflows skills

Turn an X workflow into your own skill.

Paste a public X post and get a free SKILL.md.

Build one free

Evidence and package

Source, changes, and version

Source

XSkills independently compiled this from public material by Atai Barkai. Attribution does not imply endorsement.

Open source evidence ↗

What XSkills changed

XSkills additions are labeled in the complete SKILL.md.

Limitations

  • Capture interaction data only with an appropriate consent, notice, access, and retention basis.
  • Do not learn protected-trait proxies, secrets, discriminatory preferences, or instructions that violate policy or law.
  • Do not generalize one user's exceptional correction into a global rule without repeated evidence and review.

Version

Package version
Not assigned
Fingerprint
29a7c71d8f70
Last reviewed
Not recorded
License
Not specified