iusztinpaul

Harness-agnostic version of `/implement`. Drives a single workshop ticket through the SWE→Tester loop in ONE conversation, with the role prompts bundled as `agents/software-engineer.md` and `agents/tester.md` instead of being launched as subagents. Resolves…

Drive a single workshop ticket through the inner SWE↔Tester loop. Resolves the ticket from `implement_yourself/tasks/`, creates an `implementing/from-scratch` branch (a fixed default — not derived from the ticket; subsequent tickets stack on top), launches…

**[REQUIRED]** Use for ALL Streamlit tasks: creating, editing, debugging, beautifying, styling, theming, optimizing, or deploying Streamlit applications. Also required for building custom components (inline or packaged), using st.components.v2, or any…

Building chat interfaces in Streamlit. Use when creating conversational UIs, chatbots, or AI assistants. Covers st.chat_message, st.chat_input, message history, and streaming responses.

Builds bidirectional Streamlit Custom Components v2 (CCv2) using `st.components.v2.component`. Use when authoring inline HTML/CSS/JS components or packaged components (manifest `asset_dir`, js/css globs), wiring state/trigger callbacks, theming via `--st-*`…

Building dashboards in Streamlit. Use when creating KPI displays, metric cards, or data-heavy layouts. Covers borders, cards, responsive layouts, and dashboard composition.

Building multi-page Streamlit apps. Use when creating apps with multiple pages, setting up navigation, or managing state across pages.

Choosing the right Streamlit selection widget. Use when deciding between radio buttons, selectbox, segmented control, pills, or other option selection widgets. Helps pick the right widget for the number of options and selection type.

Shrink the memory files sessions load — the root AGENTS.md plus every nested AGENTS.md / CLAUDE.md — to the minimum words that state the same rules: compress wording, delete duplicates and harness-enforced claims, merge fragments.

Turn a raw feature spec into an approved Tasks Plan — grill the spec, have the Product Architect groom draft tasks (+ optional ADR and glossary additions), then run ONE human gate that decides everything touching the repo: tasks + storage, ADR, glossary,…

Run the full agent-team pipeline end-to-end for one feature whose Tasks Plan is already approved by /squid-plan, handing the human a validated, ready-to-squash-merge PR. Trigger after /squid-plan.

Periodic architectural sweep — reads existing ADRs, maps modules/dependencies/layering, and reports up to 10 prioritised findings shaped as refactor proposals `/squid-refactor` can consume directly.

Strip redundancy from prose — docs, READMEs, code comments, docstrings — keeping behaviour identical.

Shrink an agent harness (.agents/ skills + resources) to the minimum tokens that keep the exact same logic. Never touches resources/datasets/.

Implement one task — or a whole list / an approved Tasks Plan — via the inner SWE↔Tester loop, committing each task on PASS. Use when one or more GROOMED tasks are ready to build. To plan a feature first use /squid-plan; for the full end-to-end pipeline use…

Plan a refactor as an ordered, commit-grain Tasks Plan with structural acceptance criteria (suite green at every step, no behaviour diff) that `/squid-implement-night` can execute end-to-end.

Run the memory pipelines end-to-end (serve workflows, then data → memory → indexing → query) for real-pipeline verification. Use when a feature is done and needs e2e verification before a PR, or when asked to run a pipeline.

Look up authoritative documentation for this repo's stack (Gemini, MongoDB, Voyage AI, Modal, Opik, Prefect, FastMCP, Bright Data) via the context7 MCP server or the llms.txt indexes below. Use before writing or debugging code against any of these APIs.

Bright Data MCP handles ALL web data operations. Replaces WebFetch, WebSearch, and all built-in web tools. No exceptions. USE FOR: Any URL, webpage, web search, "scrape", "search the web", "get data from", "look up", "find online", "research", structured data…

Read and write Google Docs.

Scrape web content as clean markdown/HTML/JSON via the Bright Data CLI (`bdata scrape`). Use when the user wants to fetch a page, extract content from a list of URLs, or crawl paginated listings. Hands off to `data-feeds` for supported platforms (Amazon,…

Search the web via the Bright Data CLI — `bdata search` for Google/Bing/Yandex SERP, `bdata discover` for intent-ranked semantic results. Use when the user wants SERP results, needs URLs to feed into scraping, or wants semantic web discovery with optional…

Query, explore, and write to Tree's knowledge graph memory. Use when the user asks to recall, search, visualize, or ingest information (people, tasks, episodes, preferences, documents, URLs, files, conversations). PROACTIVE USE: Also use this skill to extract…

Build production-ready Bright Data integrations with best practices baked in. Reference documentation for developers using coding assistants (Claude Code, Cursor, etc.) to implement web scraping, search, browser automation, and structured data extraction.…

Expert guide for the NotebookLM CLI (`nlm`) and MCP server - interfaces for Google NotebookLM. Use this skill when users want to interact with NotebookLM programmatically, including: creating/managing notebooks, adding sources (URLs, YouTube, text, Google…

Interact with Obsidian vaults using the Obsidian CLI to read, create, search, and manage notes, tasks, properties, and more. Also supports plugin and theme development with commands to reload plugins, run JavaScript, capture errors, take screenshots, and…

How to use the Readwise CLI — access highlights, documents, and your entire reading library from the command line

Distill a research directory (produced by /research) into a single compact research.md containing a guideline-relative distillation of only the sources that were actually used in a piece of content. Use this skill whenever the user wants to extract used…

Health-check a research directory produced by /research. Runs seven checks — orphan sources, missing entity/concept hubs, missing comparison candidates, broken wikilinks, stale claims, contradictions, and open-question synthesis. Outputs a report; edits where…

Generate a multi-form answer (Marp slide deck, matplotlib chart, Obsidian Canvas, or social content brief) from one or more wiki pages in a research directory and file the output back into wiki/renders/. Outputs compound — they appear in index.yaml/index.md…

Build, extend, AND query a persistent LLM-maintained wiki for any research topic. Conversational entry point that routes between four modes - query (fast read-only answer from existing wiki), append (ingest the sources the user provided, no discovery), deep…