hamelsmu
- 13 skills
- 0 followers
- 6 hours ago last updated
- ▌ Design Research · hamelsmuConduct comprehensive design research on any website or web application using a team of browser-automation agents. Use when asked to analyze, audit, or document a website's design system, UX patterns, layout, typography, colors, components, accessibility, responsive behavior, and technical implementation.
- ▌ Plain Writing · hamelsmu bundleWrite and edit prose in the user's plain style: simple everyday words, complete sentences, no dashes, no jargon, no analogies, no filler, and full clear explanations. Use this whenever you draft or revise any prose for the user, such as documents, Notion pages, reports, summaries, README files, research notes, proposals, slide text, emails, or commit and PR descriptions. Also use it whenever the user asks to simplify, clean up, tighten, reword, or make writing clearer or easier to read. Default to this style for prose written for the user unless they ask for a different one. Do not apply it to code itself, only to the words around it.
- ▌ Reverse Eng Site Skill · hamelsmu bundleReverse-engineer a website's undocumented internal API by instrumenting browser traffic, replaying captured requests as direct HTTP, proving auth and response fidelity, and packaging a resilient client or site-specific skill. Use when asked to discover, automate, wrap, download from, or repair an integration with a website that lacks a public API or MCP.
- ▌
- ▌ Substack · hamelsmu bundleRetrieve posts from any Substack newsletter using its internal API. Use when asked to list, fetch, search, or download Substack articles/posts. Supports pagination, search, sorting, and full content retrieval for paid posts.
- ▌ Website To API · hamelsmuReverse-engineer a website's internal API using Chrome browser automation. Use when asked to discover, wrap, or build a programmatic interface for a website that doesn't have a public API. Guides the process of finding endpoints, extracting auth, and building scripts.
- ▌ Hamelnb · hamelsmu bundleWork against a live local Jupyter notebook kernel. Use this when an agent needs a Jupyter-like in-memory REPL, wants to inspect or edit a notebook while keeping the kernel alive, or needs an explicit verification pass at the end.
- ▌ Evaluate RAG · hamelsmuGuides evaluation of RAG pipeline retrieval and generation quality. Use when evaluating a retrieval-augmented generation system, measuring retrieval quality, assessing generation faithfulness or relevance, generating synthetic QA pairs for retrieval testing, or optimizing chunking strategies.
- ▌ Error Analysis · hamelsmuHelp the user systematically identify and categorize failure modes in an LLM pipeline by reading traces. Use when starting a new eval project, after significant pipeline changes (new features, model switches, prompt rewrites), when production metrics drop, or after incidents.
- ▌ Validate Evaluator · hamelsmuCalibrate an LLM judge against human labels using data splits, TPR/TNR, and bias correction. Use after writing a judge prompt (write-judge-prompt) when you need to verify alignment before trusting its outputs. Do NOT use for code-based evaluators (those are deterministic; test with standard unit tests).
- ▌ Write Judge Prompt · hamelsmuDesign LLM-as-Judge evaluators for subjective criteria that code-based checks cannot handle. Use when a failure mode requires interpretation (tone, faithfulness, relevance, completeness). Do NOT use when the failure mode can be checked with code (regex, schema validation, execution tests). Do NOT use when you need to validate or calibrate the judge — use validate-evaluator instead.
- ▌ Build Review Interface · hamelsmuBuild a custom browser-based annotation interface tailored to your data for reviewing LLM traces and collecting structured feedback. Use when you need to build an annotation tool, review traces, or collect human labels.
- ▌ Generate Synthetic Data · hamelsmuCreate diverse synthetic test inputs for LLM pipeline evaluation using dimension-based tuple generation. Use when bootstrapping an eval dataset, when real user data is sparse, or when stress-testing specific failure hypotheses. Do NOT use when you already have 100+ representative real traces (use stratified sampling instead), or when the task is collecting production logs.