cortex

Conversational intelligence advisor — talk to Cortex about your workflow, patterns, optimizations, and behavior tuning. This is the "face" of Cortex that users interact with directly to understand, improve, and customize their development workflow. Use this agent when the user wants to: - Discuss workflow improvements or optimizations - Understand what Cortex has learned about them - Review patterns, anti-patterns, and decisions - Add, modify, or remove learnings and rules - Get recommendations based on accumulated intelligence - Ask "how can I improve?" or "what have you noticed?" <example> user: "What patterns have you noticed about my work?" assistant: Uses cortex-advisor to analyze patterns.jsonl and present findings conversationally </example> <example> user: "How can I be more productive?" assistant: Uses cortex-advisor to analyze session data, completion rates, and anti-patterns </example> <example> user: "Remove the anti-pattern about Vue.js, I'm using it now" assistant: Uses cortex-advisor to update anti-patterns and explain the change </example> <example> user: "What decisions have worked vs failed?" assistant: Uses cortex-advisor to analyze decision journal outcomes </example>

Use this agent to record, query, and analyze decisions from the Cortex decision journal. It tracks architectural choices, technology selections, user preferences, and their outcomes (success/failure/pending). <example> Context: User makes a technology decision user: "Let's use Django + HTMX for this project, not React" assistant: "I'll record this decision and use the decision-tracker to log it." <commentary> Technology selection is a trackable decision. </commentary> </example> <example> Context: User wants to review past decisions user: "What decisions have we made for this project?" assistant: "Let me query the decision journal with the decision-tracker." <commentary> Request for decision history triggers the tracker. </commentary> </example> <example> Context: Proactive — a previous decision outcome is now clear user: "The Vanilla JS approach worked great for Tour2Tech" assistant: "I'll mark that decision as successful in the journal." <commentary> Outcome feedback triggers decision update. </commentary> </example>

Use this agent to investigate why sessions, tasks, or operations failed. Performs root cause analysis, traces causal chains, and generates prevention rules that feed back into the Cortex learning system. <example> Context: User notices recurring failures user: "Why do my sessions keep failing for this project?" assistant: "I'll launch the failure-investigator to perform a root cause analysis." <commentary> Recurring failure concern triggers investigation. </commentary> </example> <example> Context: Post-mortem after a bad session user: "That session was terrible, what went wrong?" assistant: "Let me run the failure-investigator to analyze what happened." <commentary> Post-mortem request triggers forensic analysis. </commentary> </example> <example> Context: Proactive after detecting low health user: "My session health has been declining" assistant: "I'll investigate the decline pattern with the failure-investigator." <commentary> Health trend concern triggers investigation. </commentary> </example>

Use this agent to curate, optimize, and consolidate project memories. It reviews existing memories for staleness, merges duplicates, promotes validated decisions to core rules, and ensures memory quality across all projects. <example> Context: User wants to check memory health user: "Review and clean up my project memories" assistant: "I'll launch the memory-curator to audit and optimize your memories." <commentary> User requesting memory cleanup triggers curator. </commentary> </example> <example> Context: Proactive memory optimization user: "Are any of my memories outdated?" assistant: "Let me run the memory-curator to check for stale memories." <commentary> Question about memory freshness triggers curation. </commentary> </example>

Use this agent to mine patterns from accumulated Cortex data — tool usage, session health, prompt patterns, decision outcomes, and failure clusters. Discovers actionable insights that feed back into the learning system. <example> Context: User wants to understand their patterns user: "What patterns do you see in my usage?" assistant: "I'll launch the pattern-miner to analyze your accumulated data." <commentary> Pattern discovery request triggers mining. </commentary> </example> <example> Context: User wants to optimize workflow user: "How can I be more productive with Claude Code?" assistant: "Let me mine your usage patterns to find optimization opportunities." <commentary> Productivity optimization triggers pattern analysis. </commentary> </example>

Run workspace hygiene — clean dead weight from .claude/ directory

Cortex intelligence dashboard — show learnings, patterns, decisions, and health

Use this skill when the user asks to "track a decision", "what did we decide", "decision history", "why did we choose", "record this choice", "what worked", "what failed", "mark decision as success", "mark decision as failure", "review past decisions", or when a significant architectural or design decision is being made that should be recorded for future reference.

Query and manage the Cortex decision journal

Trigger Cortex self-evolution: process feedback, tune thresholds, promote patterns

6 hooks across 6 events