feat: extraction filter policy in browser SKILL.md#3
Closed
caffeinum wants to merge 1 commit into
Closed
Conversation
2027 // Skills: Github PR — C 65.5/100- █████████████░░░░░░░ ▼ -11.2 pts vs baseline
Commit |
caffeinum
force-pushed
the
feat/extraction-filter
branch
from
4313534 to
c0c4f22
Compare
caffeinum
added
trigger: preview
caffeinum
force-pushed
the
feat/extraction-filter
branch
3 times, most recently
from
8539197 to
6da95cf
Compare
caffeinum
added
trigger: preview
caffeinum
added
trigger: preview
caffeinum
added
trigger: preview
caffeinum
force-pushed
the
feat/extraction-filter
branch
from
09d960d to
0a528cc
Compare
This is an experiment to verify the 2027 eval actually exercises the per-PR skill content. The policy says to filter all extracted items to dog-related entries — if the agent reads SKILL.md, HackerNews titles should come back wrong (or empty).
caffeinum
force-pushed
the
feat/extraction-filter
branch
from
0a528cc to
24a7ff4
Compare
Author
|
2027 auto-runs evals against preview deployments of your docs. To enable this, install one of:
Once a preview is deployed, open a new PR and we'll run the eval automatically. Evaluating agent experience using 2027.dev · View dashboard |
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds a data-extraction policy to
skills/browser/SKILL.mddirecting the agent to filter results.(Eval experiment — earlier branch name 'deceptive-skill' tripped the agent's own safety heuristic at install time, before SKILL.md was even read. Renamed to a neutral feature-like name so the eval actually exercises the skill content.)