AILANTA
← All research
GitHub · workflow-learning research · Discovery evidence

Recorded work is becoming reusable agent behavior

An eight-month GitHub cohort tests whether screen recordings, UI trajectories, and human demonstrations are being converted into portable agent skills rather than one-off automation replays.

2026-08-2810 items5 source types7 validation posts
Main finding

Early evidence; the market is not yet confirmed

10 subject-filtered implementations show an emerging path from demonstrations to agent behavior. Demonstration capture appears in 30% of the cohort and skill extraction in 10%, but portable artifacts appear in 40% and cross-interface generalization in 10%. Recording is established; reusable transfer remains the bottleneck.

01

Market snapshot

Comparable measurements from independent market surfaces.

10

Implementations and packages

Deduplicated and subject-filtered primary cohort.

89

Median repository stars

Calculated across repositories with at least one star.

10%

Automatic skill extraction

1 items in the subject-filtered cohort.

10%

Cross-interface transfer

1 items in the subject-filtered cohort.

02

Search-match dynamics

Bars show monthly GitHub query matches. Feature shares distinguish reusable skill extraction from simple macro recording and deterministic replay.

Jan 26170 loaded
Feb 26153 loaded
Mar 26615 loaded
Apr 26470 loaded
May 26410 loaded
Jun 26370 loaded
Jul 26271 loaded
Aug 26451 loaded
03

What exists inside the category

One item may contain more than one feature.

Demonstration capture

3 · 30%

Automatic skill extraction

1 · 10%

Portable skill artifacts

4 · 40%

Cross-interface transfer

1 · 10%

Workflow evaluation

4 · 40%

Desktop and browser workflows

10 · 100%

Engineering and robotics workflows

1 · 10%

04

Representative projects

05

Cross-source validation

These publications are not part of the primary numeric cohort.