Fetching from the wire…
01
02
03
04
05
06
07
08
Research2026-09-15 · source-backed
Skill self-evolution methods revise skill text from execution feedback, but each oracle evaluation needs a full agent rollout, which confines search to patching whatever just failed (arXiv 2609.15396). SkillLift treats ranking as a smoother supervision target than absolute score regression and solves a bilevel problem: an inner loop revises skills against a frozen learned rubric at zero oracle cost, an outer loop spends a small number of real rollouts to re-align the rubric by rank correlation. Beats existing auto-skill methods with 40-70% less token cost than frontier evolving methods.
Each link below shares sources, entities, or timing with this story.
Skill Issue points out that auto-synthesized skill documents get optimized against tasks a capable agent already solves with no document at all, leaving the optimizer nothing to measure. The fix is mining harder tasks by reverting merged PRs at a frozen base commit. On three K...
SkillForge (arXiv 2608.24747) notes that skill-extraction approaches like SkillRL never verify whether a stored skill still works against the current environment, so the bank grows monotonically while quality rots. It makes skill usage explicit during interaction so RL optimiz...
Skill-SP couples a proposer, solver and dynamic skill controller in an RL loop where each skill guarantees deep verifiable execution in a specific scenario while routing across skills preserves variety. Environment-bound methods get precise feedback but stay narrow; open-ended...
1. Flip your multi-model pipeline to review-then-generate. Instead of using a reasoning model to plan before code generation, let the specialist generate freely and use reasoning tokens for review. Paper shows 90.2% pass@1 vs 87.2% for the planning pattern. Source 2. Audit you...
Defense-as-Skill argues pre-install vetting is structurally insufficient because a malicious skill only triggers once a concrete task and workspace state make the unsafe action look useful. SkillSonar runs as an editable skill alongside untrusted skills, checking sensitive act...
This one annoyed me, in the good way. Researchers took 206 real developer-agent sessions from 13 developers, extracted each developer's preferences from their actual interaction traces via rule-based bootstrapping plus evidence-grounded refinement, then replayed everything aga...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.