Skip to content

v0.3.25

Choose a tag to compare

View all tags
@github-actions github-actions released this 11 Aug 07:04
· 42 commits to clean-main since this release
v0.3.25
854e148 
feat: benchmark v3 stability + skills triggers in system prompt

Benchmark v3 improvements:
- Multi-run averaging (BENCHMARK_RUNS=3) with median aggregation to smooth LLM non-determinism
- Relaxed thresholds (50-150% tolerance vs old 10-20%) to account for natural LLM variance
- BENCHMARK_DISABLE=1 env var to skip checks entirely
- BENCHMARK_STRICT=1 to enable blocking mode (default: warn-only, exit 0)
- New npm script: benchmark:strict

Skills loading:
- Include trigger keywords in skill header output so agent can self-match at runtime
- buildFullContextPrompt now supports task description for auto-matching
- Updated agent prompt template to reference "Triggers" field

Version: 0.3.25
Assets 5
Loading