SkillsBench
PulseAugur coverage of SkillsBench — every cluster mentioning SkillsBench across labs, papers, and developer communities, ranked by signal.
3 day(s) with sentiment data
-
Moonshot AI releases Kimi K3.1 Pro, deliberately downgrading cyber capabilities for safety
Moonshot AI has released Kimi K3.1 Pro, a large open-weight model with a 1-million-token context window. Notably, the company deliberately removed cyber-offense training data from this model, resulting in a measurable d…
-
New method accurately measures AI agent skill internalization
Researchers have developed a new method to accurately measure when AI agents have internalized skills, allowing for more efficient retirement of redundant skills from skill hubs. Traditional methods that rely solely on …
-
Skill$^{\infty}$ enables recursive self-improvement for AI agents
Researchers have introduced Skill$^{\infty}$, a novel approach to recursive self-improvement for AI agents. This system allows a meta-agent to optimize both task-specific skills and its own meta-skill using the same age…
-
New PROVISO system enhances LLM agent security and utility
Researchers have developed PROVISO, a novel controller designed to enhance the security of Large Language Model (LLM) agents. PROVISO addresses the critical issue of LLM agents executing potentially harmful actions by e…
-
New SPEM framework boosts LLM agent skill efficiency
Researchers have introduced SPEM, a novel framework designed to enhance the efficiency and effectiveness of skill utilization in Large Language Model (LLM) agents. SPEM addresses the challenges of managing large skill l…
-
New Python tool detects LLM agent approval hijack exploits
A new Python checker tool has been developed to identify security vulnerabilities in chained LLM agent skills, specifically focusing on "approval hijacks." This technique exploits how two seemingly harmless skills can c…
-
RTK token savings claims questioned by independent AI coding cost benchmarks
A blog post from Hacker News questions the effectiveness of Rust Token Killer (RTK), a tool designed to reduce AI coding costs by compressing terminal output. While RTK claims significant token savings, independent benc…
-
New SE-GoS framework enhances LLM agent skill retrieval
Researchers have developed SE-GoS, a novel framework designed to enhance the efficiency of Large Language Model (LLM) agents by improving skill retrieval. This training-free approach evolves existing skill graphs using …
-
AI agent skills silently break during migration between platforms
A developer encountered silent failures when migrating AI agent skills from Claude Code to OpenCode, discovering that while files loaded and descriptions matched, underlying functionalities degraded. This occurred becau…
-
New AI agent methods tackle skill management and self-improvement
A new paper introduces SkillComposer, a method for managing skills in AI coding agents by treating skill selection as a joint decision rather than independent picks. This approach uses a constrained autoregressive decod…
-
LLM Agents Optimize Costs via Skill Rewriting and Translation Policies
Researchers are exploring cost-aware strategies for large language model agents to improve efficiency and performance. One paper introduces a framework for skill rewriting that optimizes for cost by preserving essential…
-
SkillDAG improves LLM agent skill selection with evolving graph
Researchers have developed SkillDAG, a novel system that models inter-skill relationships for LLM agents as a typed directed graph. This graph is dynamically updated and queried during execution, allowing agents to sele…
-
New paper identifies 'library drift' as silent failure mode in LLM skill libraries
Researchers have identified a silent failure mode in self-evolving Large Language Model (LLM) skill libraries, termed 'library drift.' This occurs when skills accumulate without proper lifecycle management, leading to d…