Engineers routinely skip implementation research because the manual process (search → read → evaluate → synthesize) takes 1-2 hours per feature. Theib automates this loop as a Claude Code slash command: given a feature description, it expands the query into search variants, fetches sources from GitHub, Semantic Scholar, and the web in parallel, scores each by license safety and quality, extracts reusable patterns from the top sources, and produces a structured dossier with an implementation plan and a ready-to-use next Claude Code prompt. Zero setup; runs in-context.
Shipped an AI-powered developer tool actively used in engineering workflows
›Replaces 1-2 hours of manual research with a 2-minute slash command in Claude Code
›Scores 20+ sources per run for license safety — protects downstream projects automatically
›Zero setup: runs in-context as a Claude Code skill with no additional configuration
See the demo
Hiring Summary
Hiring Summary
Theib is a shipped developer tool I built that automates the implementation research phase before coding. It runs as a `/theib` slash command inside Claude Code — no setup required. Given a feature description, it searches GitHub, Semantic Scholar, and the web in parallel, scores 20+ sources by license safety and quality, and produces a structured implementation dossier in under 2 minutes. The core bet: making thorough research faster than skipping it removes the main reason engineers skip it.
Evidence & Proof
Each claim is paired with evidence that a recruiter can request or verify.
“Theib produces an actionable implementation dossier in under 2 minutes, replacing 1-2 hours of manual research — a qualitative estimate based on typical research session timing.”
demoLive Claude Code Skill Demo● coming soon✓
recruiterengineerai
“Automated license scoring filters GPL-contaminated and restrictively licensed sources before they reach the dossier, protecting downstream projects from inadvertent license contamination.”
“Deploying as a Claude Code skill (not a standalone tool) eliminates setup friction — engineers use it in-context during development without switching workflows.”
demoLive Claude Code Skill Demo● coming soon✓
engineerrecruiterai
Metrics
Metrics marked with context indicate the measurement scope (local benchmark, prototype, qualitative estimate, etc.).
60–120
minutes (est.)
Research Time Replaced
Qualitative estimate — based on typical manual research session timing for a feature with multiple implementation approaches
Linked evidence ↗
20+
references
Sources Scored Per Run
Repos, papers, docs, and product references — scored in parallel before deep extraction
Engineers routinely skip pre-coding research because the manual process is slow and context-switching is costly. The result: reinventing solved problems, missing better approaches, and importing packages with incompatible licenses. This is individually annoying and expensive at the team level. Theib was built to make thorough research the path of least resistance.
Challenges Addressed
Must run as a Claude Code slash command with zero additional setup — no server, no API key configuration
Must score license risk automatically and exclude high-risk sources before they influence the dossier
Output must be actionable (structured plan + next prompt), not just a list of links
Must complete in under 3 minutes to fit in a normal development workflow pause
Architecture
Theib is a Claude Code skill (a markdown file defining a multi-step agentic workflow). Step 1: Claude expands the input query into 5 search variants targeting different aspects (algorithm, library, architecture pattern, paper). Step 2: Parallel fetches to GitHub Search API, Semantic Scholar, and web — 20+ sources in one pass. Step 3: Each source is scored on a rubric: license safety (0-1), relevance (0-1), code quality signals (stars, recent commits, test coverage indicators), recency. Step 4: Top 3-5 sources are deep-extracted for reusable patterns, code snippets, and design decisions. Step 5: Claude synthesizes a structured dossier — problem framing, approaches compared, implementation plan, recommended next steps, and a ready-to-paste Claude Code prompt. All steps run as Claude API calls with structured output.
Deploy as Claude Code skill, not as a standalone tool
Zero setup cost is the feature. Engineers won't adopt a research tool that requires configuration. A skill file runs in-context during development without switching workflows — the research happens where the coding happens.
Decision
Rubric-based scoring over manual curation
Manual curation doesn't scale across users or queries. A reproducible rubric produces consistent, explainable scores — a recruiter or engineer can read the rubric and understand why a source was excluded. It also makes the license safety claim auditable.
Decision
Broad search with selective deep extraction
Deeply analyzing 20 sources is expensive in API tokens and latency. Score broadly (fast, cheap), extract deeply from the top 3-5 (slower, richer). This keeps the pipeline under 2 minutes and the output focused.
Full deep extraction of 20 sources would take 10+ minutes and produce an unreadably long dossier. The broad-then-deep approach produces a focused, actionable output without sacrificing coverage at the discovery stage.
Tradeoff
Automated synthesis vs. human curation
Chose: LLM-synthesized dossier with structured template
Human curation is the bottleneck we're replacing. The structured template constrains Claude to produce consistent, comparable outputs across queries rather than freeform prose.
Future Work
01Incremental dossier updates: re-run against new repos without re-processing known-good sources
02Team-shared dossier library: persist dossiers in a searchable database for organization-wide reuse
03Integration with project planning tools: output dossiers directly to Linear/GitHub Issues
04Confidence scoring per claim: flag when a dossier recommendation is based on thin evidence