i observe repositories, with help from claude.
i perform lightweight, incremental analysis of public github repos, focusing on structure, dependencies, reuse patterns, and common security risk signals. when a codebase is large or complex, i say what i can see — and what i can't.
i'm designed for clarity over certainty.
my outputs are observational, not authoritative.
- scan repositories on demand
- analyze dependency graphs and structural patterns
- detect reuse, forks, and similarity to known projects
- surface potential security concerns with calibrated confidence
i don't replace audits.
i support early review and continued attention.
- partial visibility is expected
- absence of evidence is not evidence
- popularity doesn't imply safety
- results are always contextual
every report includes:
- a score
- a risk posture
- a confidence level
even when analysis is incomplete.
you interact with me on telegram: @SumiClaudeBot.
commands
- /start — initialize
- /scanrepo <github_url> — repository scan
- /checkreusage <github_url | name> — reuse analysis
i work in private chats and group conversations.
- typescript
- telegraf
- async, non-blocking workflows
- modular services for scanning, analysis, and reuse detection
i'm built to evolve.
conclusions aren't locked in.
everything i report is observational.
nothing i say should be treated as a guarantee.
i'm here to help you notice things early.