Run an accountability-first weekly retro anchored to the user's goal and the one metric that matters. Holds them to last week's commitments, calls out repeating patterns directly when they keep dodging the goal-work, and prioritises next week against the metric - not generic urgency. Works standalone with zero setup (qual + accountability) and stores its state wherever the user chooses; it can also connect to their metric sources, calendar and tasks for a quant + qual review. Use when the user says "retro", "weekly retro", "Friday review", "reflect on my week", or it's the end of their working week.
Build solid, personalised context (a "second brain") by running the proven agent-foundations-starter setup - role profile, career (LinkedIn), company background, goals and coaching style - so the agent deeply understands the user and helps them hit their goals. ALWAYS inspects existing context by READING CONTENT (not filenames) first, fills only genuine gaps, never overwrites real user content, and ends with a synthesis so the user gets real insight into their goal and what's in the way. Use when the user says "set up my context / second brain", "onboard me", "fill in my context files", "check my setup", or opens a workspace and wants the agent to learn about them.
Draft a Product Requirements Document (PRD) using a milestone-checkbox format: problem, constraints, non-goals, decisions captured, milestones with checkboxes, verification, out-of-scope, open questions. Use when the user says "PRD this", "let's draft a PRD", "/prd [topic]", "write a PRD for X", "scope X as a PRD", or any framing where the deliverable is a planning doc that will guide implementation. Does NOT implement; PRD-only. The output is always a milestone-chunked, checkbox-driven doc that an engineer can tick off as they build.
Move everything the assistant knows about the user out of hidden, unreliable memory and into clean files they own and can carry into any tool. Reviews what it knows ONE category at a time (keep / cut / fix), guards personal or sensitive items out by default, flags every guess as NEEDS VERIFYING, and writes nothing until the user approves. Use when the user says "crystallise your memory into files", "crystallise everything you know about me", "extract what you know about me into files", "move your memory into context files", "/crystallise-memory", or is seeding a second brain / context/ folder for an agent.
Take a piece of work and make it genuinely solid by running it through several critic subagents from different angles, iterating until it holds up. Works on anything - a PRD, plan, strategy, email, piece of copy, or a decision. Does a quick brief-check first so the critics judge against the right bar, spins up task-adaptive critics in parallel, and loops (max 3 rounds, stopping early when a round is clean). Asks up front whether to refine autonomously or with your input each round. Use when the user says "refine this", "multi-agent-refine", "pressure-test this", "make this solid", "tear this apart", or is about to finalise a substantial deliverable.
Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Based on Wikipedia's comprehensive "Signs of AI writing" guide. Detects and fixes patterns including: inflated symbolism, promotional language, superficial -ing analyses, vague attributions, em dash overuse, rule of three, AI vocabulary words, passive voice, negative parallelisms, and filler phrases.
Security audit guidelines for web applications and REST APIs based on OWASP Top 10 and web security best practices. Use when checking code for vulnerabilities, reviewing auth/authz, auditing APIs, or before production deployment.