Algo Sensei — DSA Mentor
A Claude-powered LeetCode & DSA mentor that teaches through progressive hints and pattern recognition instead of handing over solutions — learn to think like a senior engineer.
Pure prompt skill: no code execution, no external calls, no sensitive-data access; least privilege is inherently satisfied; ethics section explicitly discourages handing out solutions. Deducted because the publisher is unverified and the referenced modes/ files were not shown within the assessed scope, so absence of overreach downstream cannot be confirmed.
SKILL.md is internally consistent with clear mode-detection rules and few failure paths (conversational skill). But it depends on modes/*.md, templates and docs; without them functionality degrades (README admits SKILL.md-only setup lacks detailed mode instructions). No tests, no error-handling guidance; static review caps this at 10.
Audience (LeetCode/DSA learners) and per-mode trigger conditions are explicit, giving good semantic trigger precision; multi-language code support is declared. Deducted for missing non-fit boundaries, English-only documentation with no Chinese support, and full dependence on Claude Code/Claude.ai whose reachability from mainland China is uncertain.
Well-layered docs (README/SKILL/modes/templates), MIT license, contributing guide, FAQ, roadmap, and disclosed limitations (reduced quick-setup). Deducted for no versioning or changelog, vague maintenance/update ownership (generic GitHub links), and an empty placeholder scripts directory.
Core value proposition (progressive hints, Socratic teaching, pattern training) is coherent and plausibly beats directly asking for answers; output is conversational and directly usable. Deducted because no execution evidence exists: the claimed 'intelligent mode detection' and '5-level hints' are unverified, and comparative claims (vs LeetCode Premium) are marketing statements.
Only internally consistent first-party documentation and install instructions are auditable; no test suite, no CI covering key paths, no third-party validation or demos. Static cap of 5, scored down to 3.
- The skill depends on modes/, templates/ and docs/ files; copy the full directory or mode instructions will be missing and functionality degraded.
- Documentation is English-only with no Chinese support; core experience depends on Claude Code/Claude.ai reachability, which mainland-China users should verify themselves.
- No versioning or changelog; behavior may change on update—pin a specific commit when installing.
- All effectiveness claims are static documentation statements without independent execution verification; real accuracy of 'intelligent mode detection' is unknown.
- Publisher identity is unverified and the referenced modes/ files were outside this review; read all files yourself before use.
What does this skill do, and when should you use it?
Algo Sensei is a prompt-only Agent Skill that turns Claude into a data structures & algorithms mentor. It automatically routes your request into one of five modes: concept tutoring, progressive hints, code review, mock interviews, and pattern recognition training. Its teaching philosophy leans on the Socratic method and progressive disclosure — hints first, deeper reveals only when you're stuck, never a bare copy-paste solution. It supports many programming languages (Python, Java, C++, Go, and more) and tracks your skill level and language preference within a session. The MIT-licensed repository centers on SKILL.md, backed by five mode files, a solution template, and a DSA complexity cheatsheet.
It reads your problem or code, detects intent, and engages the matching mode: Tutor mode explains concepts with examples and ASCII diagrams; Hint mode delivers a 5-level progressive hint ladder (from observation nudges down to a pseudocode skeleton as a last resort); Review mode audits code for correctness, time/space complexity, edge cases, and optimization opportunities; Interview mode role-plays a professional interviewer with a performance rubric and improvement plan; Pattern Mapper mode teaches you to identify which algorithmic pattern fits a problem (two pointers, sliding window, DP, monotonic stack, etc.). Solutions follow templates/solutions/solution-template.md and reference docs/dsa-cheatsheet.md for complexity standards. It runs no scripts and makes no external calls — everything is instruction-driven.
- A job seeker prepping for technical interviews who is stuck on a problem and wants a nudge, not a spoiler, to preserve independent thinking.
- A beginner who wants concepts like dynamic programming or graph algorithms explained from the ground up at their own level.
- A LeetCode practitioner who has a working solution but wants complexity analysis and optimization feedback before moving on.
- A candidate close to a FAANG interview who wants to practice explaining solutions aloud under realistic conditions with structured feedback.
- A developer working in Python, Java, C++, or Go who wants examples and idioms adapted to their preferred language.
What are this skill's strengths and limitations?
- Free and MIT-licensed; purely prompt-based, so no dependencies or setup beyond copying files.
- Thoughtful 5-level progressive hint design that guides you to the answer without spoiling it.
- Intelligent mode routing means you describe the need instead of picking a feature manually.
- The README is candid about how it differs from LeetCode Premium, generic prompts, and video tutorials.
- The scripts/ directory is an empty placeholder for future features — there are no runnable test generators or complexity analyzers.
- Roadmap items like interactive visualizations, cross-session progress tracking, and spaced repetition are not yet implemented.
- Quality depends entirely on Claude's reasoning; the repo provides no evaluation benchmarks or test suite validating teaching outcomes.
- Installing only SKILL.md (quick setup) drops the mode detail, weakening the intended experience.
How do you install this skill?
For Claude Code (personal skills, all projects): git clone https://github.com/karanb192/algo-sensei.git, then cp -r algo-sensei ~/.claude/skills/, and restart Claude Code. For team-shared installation in a project: copy the folder to /path/to/your/project/.claude/skills/ and commit it to git. For Claude.ai: in a Project, upload SKILL.md, the 5 files from modes/, docs/dsa-cheatsheet.md, and templates/solutions/solution-template.md (a quick setup with SKILL.md alone works but omits the detailed mode instructions).
How do you use this skill?
Once installed, the skill activates automatically when you raise a DSA/LeetCode topic. Describe what you need and it picks the mode: "I'm stuck on LeetCode #3, give me a hint" triggers Hint mode; "Explain dynamic programming" triggers Tutor mode; "Review my Two Sum solution" (with code pasted) triggers Review mode; "Do a mock interview with a medium problem" triggers Interview mode; "I don't know what approach to use here" triggers Pattern Mapper mode. If you don't specify a language, it asks for your preference.
How does this skill compare with similar options?
The README itself compares it against LeetCode Premium (paid, hands you solutions), generic ChatGPT/Claude prompts (unstructured), and YouTube tutorials (non-interactive): this skill is free, DSA-specialized, and offers structured progressive learning with consistent methodology. The author notes they serve different purposes and suggests using both it and LeetCode Premium.