- Home
- Skills
- Workflow Automation
- Palm Reader: find the Claude Code skills you're missing
More screenshots
Works with the AI Agents you already use
Palm Reader: find the Claude Code skills you're missing
Reads your Claude Code history on your own machine, ranks the work you repeat, says whether each habit should be a skill, slash command, hook or CLAUDE.md line, then writes the one you pick.
$12
Palm Reader: find the Claude Code skills you're missing
Example session with this skill installed
What should I turn into skills? Look at how I've been using Claude Code.
- Read your context and instructions
- Compiled the palm reader
We'll read your Claude Code session history in ~/.claude/projects on this machine. Nothing is sent anywhere. We only quote short, redacted snippets back to you.
Palm Reader: what to turn into skills
We read
22 sessions you ran by hand, from 2026-09-21 to 2026-10-30 (70 typed prompts, split into 24 tasks), across two projects: tidepool (17 sessions) and blog (5). There were no automated sessions to leave out, so everything here is work you did yourself.
In short: Release notes are costing you the most re-explaining: you retype the same four rules almost every week, so they belong in a skill. Your screenshot resizing is the opposite: the same prompt with no follow-ups each time, so a slash command is enough. Everything else in your history was a one-off.
1. Weekly release notes: Skill
What it would do: Read the commits since the last tag in tidepool and write the week's release notes, already grouped, short, and free of ticket numbers, without you having to ask four times.
Evidence: 7 times on 7 days across 6 different weeks, between 2026-09-22 and 2026-10-30, about 145 minutes in all (roughly 20 minutes each). It took 23 follow-ups across those 7 tasks, and twice you opened with "again".
- "Write the release notes for this week from the commits since the last tag" (2026-09-22)
- "Can you do the release notes again from git log since v2.3" (2026-09-29)
- "Release notes again please, same as last time, commits since the last tag" (2026-10-27)
Rules you keep adding:
- Don't mention internal ticket numbers (like TP-123). You typed this in all 7 sessions.
- Group the changes as New, Improved and Fixed (6 sessions).
- Keep it under 200 words (6 sessions).
- Use sentence case for the headings, not title case (4 sessions).
Why a skill: It's a job with several steps (read the git log, sort the changes, write, trim) and the same four rules you keep re-explaining, and it should start on its own when you say "release notes" in your usual words.
Rough saving: about 55 minutes a month (an estimate: about 20 minutes a time, about five times a month, with half of that spent on the rules you keep re-adding).
2. Resize and compress screenshots: Slash command
What it would do: One command, say /shots shots/, that resizes the images in a folder to 1600px wide and saves compressed WebP versions, the way you ask for it now.
Evidence: 5 times on 5 days across 5 different weeks, between 2026-09-25 and 2026-10-27, about 30 minutes in all, all in the blog project. The steps were identical every time (list the folder, resize with sips, convert with cwebp), and your only follow-up was "great".
- "Resize these screenshots to 1600px wide and compress them for the blog" (2026-09-25)
- "resize the new screenshots in shots/ to 1600 wide and make webp versions" (2026-10-03)
- "resize + compress the shots folder images, 1600 wide, webp" (2026-10-27)
Why a slash command: You type roughly the same prompt each time, the steps never change and you never add a rule, so there's no judgment for a skill to carry; you just want to start it on purpose with one line.
Rough saving: about 10 minutes a month (an estimate: about 5 minutes a time, four or five times a month, a third of it saved by not writing the prompt and settling the details).
Want us to write one of these? Pick a number.
(Palm Reader, by Figmatro)
Connects securely to your tools. The creator never sees your data.
About this skill
You already know what your skills should be. They're the things you keep asking Claude to do, with the same corrections every time: "keep it under 200 words", "don't mention ticket numbers", "run the tests first". Palm Reader reads the lines in your Claude Code history and tells you which skills are in your future. Then it writes the one you pick.
How it works
- Ask Claude. "What should I turn into skills?" or "read my palm".
- Palm Reader reads your history, on your machine, in seconds. It leaves out automated runs, splits long sessions into separate jobs, and groups the work you repeat.
- You get a reading, ranked. Each habit comes with the evidence (how often, over how many weeks, and a few of your own prompts), the rules you keep adding, and the right size for it.
- It writes the one you pick. You get a skill whose description uses the words you actually type, rules taken from your corrections, and test prompts from your real requests. Or you get a slash command, a hook, or a fix to a skill you already have. Nothing is installed without your yes.
The right size for every habit
- Skill: multi-step work with rules you keep re-explaining, like weekly release notes or PR reviews.
- Slash command: a prompt you type the same way every time.
- Hook: something that must happen every time, like tests before a commit. A skill only runs when it triggers, and a hook always does.
- CLAUDE.md line: a single preference you keep restating.
- Description fix: a skill you already have that never triggers.
Palm Reader never invents a habit. If your history is too short to say much, it says so and shows you how to keep more.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Workflow Automation.
Claude Code only: claude.ai chats aren't stored locally. Needs a few weeks of regular use to spot weekly habits; with less it says there isn't enough yet. Claude Code deletes history after 30 days by default (Palm Reader shows you how to keep more). Headless runs, evals, scheduled tasks and subagents are left out on purpose.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 13 days ago
- Software
- Claude Code
Python 3.8+
Git
Reads Claude Code's local history in ~/.claude/projects.
Permissions
Terminal: runs one Python script that reads your Claude Code session files on your own machine. Read files: your session history and installed skills, to find habits and avoid duplicates. Write files: a short local summary, with keys, tokens and email addresses taken out, and the skill, command or hook you pick, only after you say yes. No network access: nothing is sent anywhere.