Agent skill

paste

Quellcode ansehen: coreyhaines31/makerskills

#201Globales Ranking · von 201 SkillsSafe

Installation

npx skills add coreyhaines31/makerskills --skill paste

269

Installationen

EU-hosted inference API

Power your AI agent skills with open-source models.

Drop-in OpenAI-compatible API. No data leaves Europe.

MiniMax

MiniMax M3

$0.40 / $1.40

per M tokens

Z.ai

GLM 5.3 Flash

$0.20 / $0.60

per M tokens

MoonshotAI

Kimi K3

$4.00 / $18.00

per M tokens

DeepSeek

DeepSeek V4.1 Flash

$0.40 / $1.40

per M tokens

/paste — Clean content for any destination

Cleans terminal output (ANSI, box-drawing, prompt artifacts, etc.) and reformats per destination rules. Reads clipboard by default, writes back to clipboard + previews in chat.

Step 1 — Get the content

In order:

  1. If the user included content in the prompt (pasted, or referenced from earlier in the conversation), use that.
  2. Else, read clipboard: pbpaste
  3. If both empty, ask the user what to paste.

Step 2 — Parse destination

Invocation Destination
/paste plain (default)
/paste plain plain
/paste slack slack
/paste notion notion
/paste twitter or /paste x twitter
/paste linkedin or /paste li linkedin
/paste email email (plain)
/paste email rich or /paste email html email (rich/HTML)
/paste github or /paste gh github
/paste markdown or /paste md markdown render
/paste html render to HTML file + open in browser

If ambiguous, ask. Per-destination rules live in references/destinations.md.

Step 3 — Scan for secrets

Before any cleaning or copying, scan against patterns in references/secret-patterns.md.

If any match:

  1. Stop. Don't copy yet.
  2. Report what was detected with the value masked (first 4 + last 4 chars only — e.g. sk-12...wxyz).
  3. Ask: "Spotted [type] in the content. Continue, redact (replace with [REDACTED]), or abort?"
  4. Default to abort if no clear answer.

Step 4 — Universal cleaning

Strip from the content:

  • ANSI escape codes\x1b\[[0-9;]*[a-zA-Z] and related (color codes, cursor movement, screen control)
  • Box-drawing chars╭ ─ ╮ │ ╰ ╯ ╞ ╡ ╤ ╧ ┌ ┐ └ ┘ ├ ┤ ┬ ┴ ┼ ━ ┃ ┏ ┓ ┗ ┛ etc. Replace with simpler equivalents (-, |) or remove entirely depending on context.
  • Terminal prompt artifacts — leading $ , > , , % , # (when first non-whitespace on a line that looks like a shell prompt)
  • Carriage returns (\r) — replace with \n or delete
  • Trailing whitespace per line
  • Excess blank lines — collapse 3+ consecutive blank lines to 2

Step 5 — Apply destination rules

Read references/destinations.md and apply the relevant transform.

Step 6 — Output

Destination Output behavior
plain / slack / notion / twitter / linkedin / email / github (a) Show preview in a code fence in chat. (b) Copy clean version to clipboard via pbcopy.
email rich Render to HTML, write to /tmp/paste-<timestamp>.html, open it in browser. Skip clipboard (the user copies from browser to preserve rich text).
markdown Render the cleaned markdown directly in chat (Claude Code renders it). Also copy raw markdown to clipboard. Offer: "Open as HTML too?" — if yes, write and open.
html Render to HTML, write to /tmp/paste-<timestamp>.html, open it. Skip clipboard.

After output, report a one-line summary of what was cleaned (e.g., "Stripped 12 ANSI codes, 4 box-drawing chars, 1 prompt artifact. Character count: 248 / 280 (X).").

Length warnings

Twitter and LinkedIn have practical limits. the user said "let me trim" — don't refuse or auto-truncate.

  • Twitter/X: warn if >280 chars. Note the overage.
  • LinkedIn: warn if >400 chars (comfort line, not hard limit).
  • Other destinations: no length warning.

URL handling for social destinations

If the content has URLs and destination is twitter or linkedin:

  • Surface the URLs separately (don't include them in the body)
  • Remind the user: links go in a first comment (LinkedIn) or reply (X), not the body
  • See makerskills:jab-hook and ~/.claude/memory/feedback_social_link_placement.md

Examples

/paste slack

→ Reads clipboard, strips ANSI, converts **bold** to *bold*, copies to clipboard, shows preview.

/paste twitter
[pasted error log]

→ Uses pasted content (ignores clipboard), strips everything, warns at 412/280 chars, copies, asks if the user wants to trim or pick a different destination.

/paste html
[pasted markdown table]

→ Renders table as HTML, opens in browser. the user selects + copies into Notion/email with formatting preserved.

/paste
[pasted output with sk-anthropic-key in it]

→ Stops. "Spotted what looks like an Anthropic API key: sk-an...3kf9. Continue, redact, or abort?"

Composes with

  • jab-hook — clean output for pasting into Typefully drafts when the MCP path doesn't fit; also enforces the link-placement rule for X/LinkedIn destinations
  • social-fetch — clean a fetched post before quoting it in a draft or newsletter
  • second-brainpaste output can be captured cleanly into raw/note-<slug>.md for future compilation
  • watch-video — transcript / summary output often flows through paste for platform-specific reformatting

Notes on quality

  • Secret detection runs first, always. Before any formatting or destination logic, paste scans for sk-*, AKIA*, xoxb-*, JWT-shaped strings, and long-hex tokens. Prompts before proceeding on any hit. Better to false-positive occasionally than to leak a real key.
  • Destination-aware transforms, not one-size-fits-all. Slack wants *bold*; LinkedIn wants unicode-styled bold; email wants HTML. Same input, 9 different valid outputs. The destination flag is not optional.
  • ANSI codes get stripped for every destination. Terminal escapes (\033[31m etc.) render as garbage everywhere except the source terminal.
  • Character limits are enforced, not warnings. X at 280, LinkedIn at 3000, Twitter at 25000 — paste refuses to copy over-limit output and offers a trim strategy.
  • Links go in first comments for social destinations — not the body. Enforced when destination is twitter or linkedin. Documented in ~/.claude/memory/feedback_social_link_placement.md.
  • HTML destination opens in a browser tab, not the clipboard. Formatted tables + code blocks need a rendered surface to select-and-copy from with formatting preserved. Copying raw HTML to the clipboard produces garbage in Notion/email.

Installationen

Installationen269
Globales Ranking#201 von 201

Sicherheitsprüfung

athSafe
socketSafe
Warnungen: 0Bewertung: 90
snykLow
WEB DATA FOR AGENTS

Give agents clean web context

Search and extract the public web as Markdown or structured JSON through one API or hosted MCP server.

Explore Webstractor

So verwenden Sie diesen Skill

1

Install paste by running npx skills add coreyhaines31/makerskills --skill paste in your project directory. Führen Sie den obigen Installationsbefehl in Ihrem Projektverzeichnis aus. Die Skill-Datei wird von GitHub heruntergeladen und in Ihrem Projekt platziert.

2

Keine Konfiguration erforderlich. Ihr KI-Agent (Claude Code, Cursor, Windsurf usw.) erkennt installierte Skills automatisch und nutzt sie als Kontext bei der Code-Generierung.

3

Der Skill verbessert das Verständnis Ihres Agenten für paste, und hilft ihm, etablierte Muster zu befolgen, häufige Fehler zu vermeiden und produktionsreifen Code zu erzeugen.

Was Sie erhalten

Skills sind Klartext-Anweisungsdateien — kein ausführbarer Code. Sie kodieren Expertenwissen über Frameworks, Sprachen oder Tools, das Ihr KI-Agent liest, um seine Ausgabe zu verbessern. Das bedeutet null Laufzeit-Overhead, keine Abhängigkeitskonflikte und volle Transparenz: Sie können jede Anweisung vor der Installation lesen und prüfen.

Kompatibilität

Dieser Skill funktioniert mit jedem KI-Coding-Agenten, der das skills.sh-Format unterstützt, einschließlich Claude Code (Anthropic), Cursor, Windsurf, Cline, Aider und anderen Tools, die projektbezogene Kontextdateien lesen. Skills sind auf Transportebene framework-agnostisch — der Inhalt bestimmt, für welche Sprache oder welches Framework er gilt.

Data sourced from the skills.sh registry and GitHub. Install counts and security audits are updated regularly.