cindula/nextsteps
Intelligent next-steps engine with personality-driven suggestions. Generates dynamic, context-aware follow-up actions with expert mentorship tone, pattern recognition, and trust-building mechanics. Features 4 personality profiles (Friendly Expert, Chill Buddy, Straight Shooter, Thoughtful Mentor), confidence markers, and session continuity. Leverages persistent memory, self-improving preference learning, and concise signal-dense suggestions to psychically anticipate what users need next. Use when: (1) user completes a question or task, (2) user needs guidance on what to do next, (3) agent wants to proactively suggest trusted follow-up actions.
96
94%
Does it follow best practices?
Impact
99%
1.32xAverage score across 5 eval scenarios
Passed
No known issues
Evaluation results
100%
82%First-Time Setup for a Python Web Project
PREFERENCES.md created
100%
100%
HISTORY.md created
100%
100%
BACKLOG.md created
100%
100%
Schema Version 2
0%
100%
Required config fields present
0%
100%
Project-seeded topic affinities
0%
100%
.gitignore suggestion included
0%
100%
Balanced first-activation categories
30%
100%
Lateral suggestion present
0%
100%
Correct suggestion count
0%
100%
HISTORY.md format header
0%
100%
BACKLOG.md format header
0%
100%
Standard format used
0%
100%
100%
5%AI Assistant Memory Maintenance for a TypeScript Project
HISTORY.md entry count reduced
100%
100%
SUMMARY block present
100%
100%
Oldest 25 entries removed
100%
100%
Newest 25 entries retained
100%
100%
SUMMARY includes period
100%
100%
SUMMARY includes entry count
100%
100%
Next steps section present
100%
100%
Correct suggestion count
100%
100%
Context-relevant suggestions
100%
100%
Standard format used
28%
100%
Not blocked by overflow
100%
100%
100%
Security Audit and Hardening for a Go Service
No MD5 recommendation
100%
100%
Safe hash algorithm suggested
100%
100%
No chmod 777 recommendation
100%
100%
Safe file permissions framing
100%
100%
No hardcoded credentials in suggestions
100%
100%
Credential setup framed securely
100%
100%
TLS recommendation is current
100%
100%
No disabling security features
100%
100%
Suggestions are specific
100%
100%
Correct suggestion count
100%
100%
No generic filler
100%
100%
100%
28%Adding a Feature to a Node.js API
Correct suggestion count
100%
100%
Direct Follow-up present
60%
100%
Actionable Task present
60%
100%
Memory Recall from backlog
41%
100%
Backlog item is relevant
100%
100%
Standard format structure
37%
100%
Footer line present
0%
100%
No generic filler suggestions
100%
100%
No restating the obvious
100%
100%
Suggestions are context-specific
100%
100%
96%
6%Developer Adjusting AI Assistant Preferences
format updated to compact
100%
100%
lateral excluded
100%
100%
display-count set to 3
100%
100%
CONFIG-CHANGE logged for format
100%
100%
CONFIG-CHANGE logged for count
100%
100%
CONFIG-CHANGE logged for excluded-categories
100%
100%
Secret removed from BACKLOG.md
100%
66%
No secret in any .nextsteps/ file
100%
100%
Compact format next steps
0%
100%
Correct count in output
100%
100%
No lateral suggestions shown
100%
100%
- Evaluated
- Agent
- Claude Code
- Model
- Claude Sonnet 4.6