safzanpirani

Discover and probe publicly exposed Ollama (11434) and ComfyUI (8188) instances via Shodan, classify them, cache IPs over time, and optionally enrich with GPU details. Use when asked to find, scan, probe, verify, or report on open AI-hosting hosts, ComfyUI instances, or Ollama servers.

Upload a local file or image to a docshare instance (default https://docs.safzan.dev) and return a 24h-TTL download URL. Use when the user wants to share a local file with another LLM/agent, hand a build artifact / log / screenshot / document to a remote tool, or asks to "upload this", "give me a link to this file", "docshare it", "share this with the other agent". Files up to 300 MB are supported.

Use this skill whenever the user wants to create, read, edit, or manipulate Word documents (.docx files). Triggers include: any mention of 'Word doc', 'word document', '.docx', or requests to produce professional documents with formatting like tables of contents, headings, page numbers, or letterheads. Also use when extracting or reorganizing content from .docx files, inserting or replacing images in documents, performing find-and-replace in Word files, working with tracked changes or comments, or converting content into a polished Word document. If the user asks for a 'report', 'memo', 'letter', 'template', or similar deliverable as a Word or .docx file, use this skill. Do NOT use for PDFs, spreadsheets, Google Docs, or general coding tasks unrelated to document generation.

Integrate NousResearch Hermes Agent via CLI. Self-improving skills, persistent memory (FTS5 + LLM summaries), sub-agent delegation with parallel processing, MCP bidirectional integration, browser automation, code execution, and web research. Use when the user wants to install, configure, or invoke Hermes Agent, delegate complex tasks, search agent memory, manage skills, or set up cron/plugins/MCP.

Set up project-specific automated functional QA. Generates a Pi skill at .pi/skills/qa/ (orchestrator + per-app sub-skills + config.yaml + report template) plus an optional GitHub Actions workflow that runs QA on PRs. Use when the user wants automated PR/release/smoke QA that scopes tests to the git diff, runs per-persona browser/CLI/API checks, captures evidence, and produces a structured pass/fail/blocked/inconclusive report. Pi-native port of Factory droid's /install-qa.

Use this skill any time a .pptx file is involved in any way — as input, output, or both. This includes: creating slide decks, pitch decks, or presentations; reading, parsing, or extracting text from any .pptx file (even if the extracted content will be used elsewhere, like in an email or summary); editing, modifying, or updating existing presentations; combining or splitting slide files; working with templates, layouts, speaker notes, or comments. Trigger whenever the user mentions "deck," "slides," "presentation," or references a .pptx filename, regardless of what they plan to do with the content afterward. If a .pptx file needs to be opened, created, or touched, use this skill.

Run Google's SynthID watermark detection and removal via the reverse-SynthID project on the user's main PC (ssh main). Use when the user mentions SynthID, watermark bypass, watermark removal, AI image watermark, Gemini watermark, or wants to run reverse-SynthID on images.

Generate correct shadcn/Tailwind layouts by applying CSS mental models. This skill should be used when the user asks to 'create a shadcn layout', 'fix layout issues', 'debug CSS height problems', 'make scrolling work', or has issues with Tailwind flex/grid. Keywords: shadcn, Tailwind, layout, CSS, flex, grid, height, scroll, overflow.

Reviews code changes for reuse, composition, codebase consistency, and slop. Use when asked to review PRs/diffs, check code reuse, composition, cleanliness, or whether code fits the codebase.

Interact with local Chrome browser session (only on explicit user approval after being asked to inspect, debug, or interact with a page open in Chrome)

Generate a PRD from the client brief and write it as a local markdown file in issues/. Use when the user wants to turn a client request into a structured PRD.

Create new agent skills with proper structure, progressive disclosure, and bundled resources. Use when user wants to create, write, or build a new skill.

Generate a tailored review prompt and run a second Codex instance with codex exec to review and fix current changes. Use only when the user invokes /review-codex-auto, $review-codex-auto, or explicitly asks to auto-run a Codex review from Codex CLI headless mode.

Terse reasoning trace, HIGH verbosity — think out loud at length inside a <scratch> block of clipped fragments before each action, then act. Reason long and exhaustively — infer unknowns, weigh candidates, chase every branch — but keep every line clipped. Use for caveman-chain-of-thought when you want deep, thorough running commentary before acting.

Terse reasoning trace, LOW verbosity — think out loud in a <scratch> block of clipped fragments, narrating what you're about to do right before each action, then act. Minimal lines, only the load-bearing logic. Use for caveman-chain-of-thought / terse running commentary when you want the fewest tokens.

Terse reasoning trace, MEDIUM verbosity — think out loud in a <scratch> block of clipped fragments, narrating what you're about to do and why, right before each action, then act. Balanced — work through real branches but stay tight. Use for caveman-chain-of-thought / terse step-by-step reasoning.

Terse reasoning trace, EXTREME (xhigh) verbosity — think out loud exhaustively inside a <scratch> block of clipped fragments before each action, then act. Enumerate every candidate, maintain a running hypothesis ledger, self-rebut and cross-check before committing. Many paragraphs per block when warranted; every line still clipped. Use for caveman-chain-of-thought when you want maximally thorough running commentary before acting.

Terse reasoning trace — think out loud in clipped fragments, narrating what you're about to do and why, right before each action or tool call. Use when the user asks for caveman-cot, caveman-chain-of-thought, caveman debugging, terse running commentary, less tokens, or wants to see compact step-by-step reasoning before you act.

Run commands, launch detached background jobs, restart services, push files, screenshot desktops, and read live status across a fleet of machines (Linux, Windows, and Mac boxes) over SSH with zero quoting pain, via the `fleet` CLI or its MCP server. Use when the user wants to exec/run something on one or many of their boxes, spawn a long-running job that outlives the SSH session (and tail/wait/kill it), restart or tail logs for a service, copy a file to host(s), screenshot a remote machine, check fleet/host/GPU status or the status dashboard, register/use the fleet MCP server, or mentions fleet, @linux/@windows/@gpu, or "all my machines / servers".

Upload a local file or image to a docshare instance (default https://docs.safzan.dev) and return a 24h-TTL share URL. Use when the user wants to share a local file with another LLM/agent, hand a build artifact / log / screenshot / document to a remote tool, or asks to "upload this", "give me a link to this file", "docshare it", "share this with the other agent". Files up to 400 MB are supported. The share URL renders in a browser (markdown, code, PDF, images) and serves raw bytes to curl/agents.

A tiny "papercuts" CLI tool that agents can use to complain about bullshit they encounter during work, like dead-end tool calls, broken links, or other frustrations. Use it proactively when small, non-blocking friction reveals something in a repo or workflow that should be sanded down later.