Linked Agency | Ghostwriting & DFY LinkedIn

Anthropic cut 80% of Claude’s system prompt. It got better (yours can too). Boris Cherny, who built Claude Code: "Every six months, delete your CLAUDE.md. Delete your skills. Delete your hooks." I'm not deleting mine. Do this instead → https://lnkd.in/eeND5P4t Repost ♻️ to help just one person in your network. P.S. Are you still running Claude from 6 months ago?

I've hoarded 100+ Claude repos this year. These are my top 30 (all free + links below): HARNESS, what it does before it types. ✦ learn-claude-code → https://lnkd.in/eyJE44vM ✦ karpathy-skills → https://lnkd.in/eimJx6dS ✦ superpowers → https://lnkd.in/eppbgRaK ✦ ponytail → https://lnkd.in/eYzPcnAR ✦ gstack → github.com/garrytan/gstack ✦ ECC → github.com/affaan-m/ECC Fewer rewrites, because it plans before it types. SKILLS, the jobs it repeats. ✦ taste-skill → https://lnkd.in/eEZn2dx4 ✦ anthropics/skills → https://lnkd.in/er2tG4ZB ✦ wshobson/agents → github.com/wshobson/agents ✦ claude-plugins → https://lnkd.in/e6_tjwuP ✦ ui-ux-pro-max → https://lnkd.in/eh5s-T6t ✦ awesome-claude-skills → https://lnkd.in/eksG64au This is what kills the generic AI look. MEMORY, what survives the session. ✦ planning-with-files → https://lnkd.in/ev7S5cyG ✦ claude-mem → https://lnkd.in/e_AcBv2A ✦ codegraph → https://lnkd.in/eH83_uT9 ✦ graphify → https://lnkd.in/eT9f97nW ✦ repomix → https://lnkd.in/evuAbe6s It stops re-reading your whole repo every session. TOOLS, what it reaches past the terminal. ✦ multica → https://lnkd.in/eC4G4yQY ✦ firecrawl → https://lnkd.in/eKhyjZUk ✦ cc-switch → https://lnkd.in/ehCWJyYA ✦ vibe-kanban → https://lnkd.in/erSYesUg ✦ github-mcp → https://lnkd.in/eK9RtiNK ✦ playwright-mcp → https://lnkd.in/eeDRfZny ✦ claude-code-router → https://lnkd.in/enVRpRmU ✦ awesome-mcp-servers → https://lnkd.in/ebE2iDvV Now it tests its own work in a real browser. COST, what it actually burns. ✦ system-prompts-ai → https://lnkd.in/ed7gQn2J ✦ best-practice → https://lnkd.in/eSY5347M ✦ codex-plugin-cc → https://lnkd.in/eTweEPmw ✦ claude-hud → https://lnkd.in/ecU6_AC9 ✦ caveman → https://lnkd.in/e4nxpEJi Same answers. Far fewer tokens. New to Claude? Start here → https://lnkd.in/eM89acTy Repost ♻️ to help someone in your network. P.S. Which one are you installing first?

.md, Skills, Hooks, Subagents, MCP. Five words everyone uses and nobody explains: Full guide → https://lnkd.in/eeND5P4t CLAUDE.MD, the brief it reads before you type. ✦ Loads every session ✦ Mine says "always plan first" ✦ Run /init and it writes you a starter It guides, though. It cannot force anything, which is exactly why hooks exist. SKILLS, the jobs you do on repeat. ✦ Fires when you call it, or when it fits ✦ /code-review instead of a 200-word prompt ✦ Add a SKILL.md in .claude/skills/ It costs you nothing sitting there unused, because the detail only loads when it actually runs. HOOKS, the rule that fires without you. ✦ Runs always, at the moment you picked ✦ Tidies every file the second it saves ✦ Run /hooks and pick the moment Now one fires on every prompt I send. SUBAGENTS, the specialist you hand one job. ✦ Its own head, its own memory ✦ A researcher that reads but can't edit ✦ Add a .md file in .claude/agents/ It can't see your chat, so put the whole brief in writing and it'll never eat into your context. MCP, the socket to the apps you already pay for. ✦ Runs the moment Claude needs live data ✦ Mine writes straight into my Notion ✦ claude mcp add, then check with /mcp Whatever you connect, Claude can read, so only wire up the apps you already trust. The difference is WHEN each one runs. CLAUDE.md guides. A hook forces. That's the bit that took me longest to get. Every prompt and skill → https://lnkd.in/eM89acTy More free AI guides → https://lnkd.in/eKtw2V6e Repost ♻️ to help someone in your network. P.S. Which of the five did you set up last?

Claude can (finally) read every meeting you've had. 12 combos that turn a call into finished work: All 12 run on one input (the notetaker). If that's wrong, everything breaks across the system. E.g. my team is Smriti, Jonnibet and Ansh. Other notetakers hand me back Smitty, Jonnibat and Ant. - A wrong name → nobody owns the ticket - A dropped "not" → a promise you refused - "4th" heard as "14th" → you miss the call - "MCP" heard as "MCB" → means nothing Mostly right is still wrong. That's the part Wispr Flow Notetaker is built around. - Works on any platforms (even Slack huddles) - No meeting bots join your work calls. - Terms come out of your dictionary. - Names come off the invite. Then its MCP hands that transcript to Claude, and Claude's connectors do the rest. Some of my favourite example tool combos: ✦ Claude → total recall "What did everyone commit to this week? Cite each one." ✦ Slack → a decision log "Post today's decisions to the SOC 2 channel with owners and dates." ✦ Notion → a wiki that writes itself "File this week's summaries into the project page, keep the links." ✦ HubSpot → a CRM that updates itself "Log yesterday's calls onto the right deal with the next step." ✦ Drive → a pattern finder "Pull every pricing objection this quarter into one doc, by theme." ✦ Calendar → follow-ups already booked "Book every follow-up we agreed, with the right people on it." ✦ Linear → tickets that open themselves "Open an issue per commitment, assigned to whoever said it." ✦ Asana → a task list with owners "Turn every action from today's calls into tasks, with due dates." ✦ Airtable → a commitment tracker "Add every promise made this week to the tracker base." ✦ PandaDoc → a proposal first draft "Draft the proposal from Tuesday's scoping call, in their words." ✦ Canva → a recap deck "Turn the QBR call into a six-slide recap deck for the client." ✦ Zapier → everything downstream "Fire the onboarding zap for every client who said yes today." Wire up the two or three you actually live in. Three connectors you use will beat twelve you don't. Notetaker is free to use → https://lnkd.in/daRajTHy Code CHARLIE gets you a month of Pro. Desktop only, Settings → Plans and Billing. Repost ♻️ to help someone in your network. #WisprFlowPartner

Claude Code has 118 commands. Here are the 15 worth learning first: ✦ /init - Generates your CLAUDE.md automatically - Claude reads your files and writes the brief ✦ /plan - Switches to read-only mode - Maps every step before touching a file - Five minutes of planning saves an hour ✦ /context - Shows what is eating your tokens - Most people are 15-20% through before typing ✦ /compact - Frees up context mid-session - Run this at 50% usage, not 95% ✦ /clear - Full reset between tasks - Stops old context degrading new output ✦ /model - Switch Opus, Sonnet, Haiku mid-session - Opus for depth. Haiku for speed. ✦ /goal - Set the finish line before Claude starts - Claude works until the condition is met ✦ /rewind - Undo when Claude goes off track - Or hit Esc twice for the rewind menu ✦ /agents - Create and manage sub-agents - Split big tasks across parallel workers ✦ /workflows - Claude splits the job across a team of agents - Watch, pause and steer mid-run ✦ /effort - One dial for thinking depth - Low for speed. Max for hard problems. ✦ /simplify - Three agents review your code in parallel - Catches architecture issues and duplicates ✦ /permissions - Pre-allow safe commands instead of bypass ✦ Shift+Tab - Cycles default, accept edits, plan and auto mode ✦ "ultracode" - One keyword. Claude plans a full agent workflow. This guide has all 118 across 10 categories. Save it → https://lnkd.in/eM89acTy Repost ♻️ to help someone in your network. P.S. Which one are you using the most?

Claude can do (almost) any job for you. You only need to install the right skill. A skill is a folder of instructions Claude loads on demand. It turns Claude into a specialist with zero prompting from you. Every skill installs with one command: ✦ Find Skills (1.9M installs) Finds Claude skills for you → https://lnkd.in/eFcmubhq ✦ Agent Browser (432K installs) Lets Claude browse for you → https://lnkd.in/eUS4cxjs ✦ Superpowers (211K installs) Plan, test and debug code → https://lnkd.in/eppbgRaK ✦ Caveman (142K installs) Cut tokens by ~75% → https://lnkd.in/e3uPsgtM ✦ Frontend Design (521K installs) Anthropic's frontend skill → https://lnkd.in/er2tG4ZB ✦ Design Taste (128K installs) Makes frontend premium → https://lnkd.in/emKJrHrX ✦ Impeccable Design (54K installs) Top-tier frontend polish → https://lnkd.in/eMSX587R ✦ Remotion (359K installs) Builds full videos in React → https://lnkd.in/ecX3gdvr ✦ Website to HyperFrames (83K installs) Creates videos in HTML → https://lnkd.in/ed-wSdsx ✦ GSAP (19K installs) Smooth web animations → https://lnkd.in/e6avMtKZ ✦ Marketing Skills (132K installs) SEO, ads and copy → https://lnkd.in/egt-7ZwM ✦ Social Media Skills (1.4K GitHub stars) My content skills pack → https://lnkd.in/emDvetxm I hand-picked the 53 best, grouped into 5 categories. The full list is here → https://lnkd.in/eM89acTy You are not learning 53 new tools. You are teaching one tool 53 new jobs. Repost ♻️ to help someone in your network. P.S. Which skill are you installing first?

Anthropic deleted 80% of Claude instructions. It got better. I built a (free) skill that updates yours: "Every six months, delete your CLAUDE.md. Delete your skills. Delete your hooks." That's exactly what Boris Cherny (creator of Claude Code) recently said about the latest Opus 5 model. Because every instruction in your setup was written to fix a problem an older model had. The model moved. Your instructions should too: - It reads Anthropic's live docs for your model. - It checks your CLAUDE.md, skills and hooks. - It tells you every line: delete, keep, rewrite. - It hands you the implementation plan. You just have to read and approve it. Grab the skill here → https://lnkd.in/eM89acTy Repost ♻️ to help someone in your network. P.S. When did you last update Claude?