feat: add anchor-drift-monitor skill - real-time intent drift detection for Oz runs#16
Open
HasRahm wants to merge 1 commit into
Open
feat: add anchor-drift-monitor skill - real-time intent drift detection for Oz runs#16HasRahm wants to merge 1 commit into
HasRahm wants to merge 1 commit into
Conversation
Add documentation for the Anchor Drift Monitor skill, detailing its purpose, usage, prerequisites, workflow, and scoring guidelines.
2 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Adds a new
anchor-drift-monitorskill that detects and prevents cumulative intent drift during multi-step Oz agent runs.Why
Oz agents can experience cumulative intent drift across multi-step tasks. An agent starts with a clear instruction - "refactor the auth module" - and by step 6 it is modifying unrelated files, changing environment variables, or introducing dependencies outside the original scope. This happens because multi-step context causes gradual faithfulness erosion.
How It Works
Benchmark Results
Validated on CNN/DailyMail corpus using Gemma 4 31B + HHEM-2.1-Open:
3:1 improvement ratio - interventions are grounded, not random.
0% regression on IFEval (instruction following) and Tau2 (task completion).
Setup
Requires
ANCHOR_KEYenvironment variable. Free keys at anchor-app-one.vercel.app.Links
npm install @anchor-engine/sdk