synapse

Synapse acts as the central intelligence layer within the EvolveSkill ecosystem, designed to bridge the gap between historical project data and autonomous agent improvement. By continuously observing agent activity, Cortex memory logs, and technical documentation such as PRD files or roadmap markdown documents, Synapse identifies recurring engineering patterns. When it detects frequent technical tasks—such as repetitive API integrations, boilerplate testing setups, or deployment configurations—it automatically triggers the creation of new, reusable skills. This allows the AI agent to transition from generic responses to a highly personalized workflow that adapts to the specific language, framework, and architectural choices of the codebase. It is intended for software engineers and autonomous agents operating in long-term development environments who require persistent, evolving toolsets. The system functions by evaluating patterns against a configurable threshold, ensuring that only high-impact, frequently occurring tasks are promoted into standalone automated modules. Users simply code normally, and the system handles the identification and recommendation process without manual intervention.

• Automated discovery of recurring work patterns in .cortex_log.md, task lists, and PRD files.

• Intelligent priority scoring (Critical, High, Medium, Low) based on frequency of occurrence.

• Seamless integration with the Cortex API for persistent event tracking across multiple AI sessions.

• Generation of structured, portable skills that follow progressive disclosure and best practices for Claude Code, GPT, and Gemini CLI agents.

• Configurable analysis windows and sensitivity thresholds via .synapse_config.json to tailor recommendations to specific project needs.

• Native support for git hooks, ensuring that pattern detection occurs at every commit for real-time observability.

• Synapse periodically scans project directories to parse checkboxes, lists, and task requirements to infer missing capabilities.

• The engine generates a prioritized Synapse_RECOMMENDATIONS.md file containing full technical rationales and proposed skill logic.

• Once the frequency threshold (default 5 occurrences) is met, the system generates a ready-to-use directory containing SKILL.md and automation scripts.

• Ensure the repository is configured with proper Cortex hooks to enable maximum context capture.

• Use command-line arguments to trigger standalone analysis for manual review before allowing automated skill generation.

• The system works best when paired with Forge for immediate script generation and MCP Provider for external tool connectivity.