Agent skill
paste
Quellcode ansehen: coreyhaines31/makerskills
Installation
npx skills add coreyhaines31/makerskills --skill paste269
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:
- If the user included content in the prompt (pasted, or referenced from earlier in the conversation), use that.
- Else, read clipboard:
pbpaste - 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 |
|
/paste linkedin or /paste li |
|
/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:
- Stop. Don't copy yet.
- Report what was detected with the value masked (first 4 + last 4 chars only — e.g.
sk-12...wxyz). - Ask: "Spotted [type] in the content. Continue, redact (replace with
[REDACTED]), or abort?" - 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\nor 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-hookand~/.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 destinationssocial-fetch— clean a fetched post before quoting it in a draft or newslettersecond-brain—pasteoutput can be captured cleanly intoraw/note-<slug>.mdfor future compilationwatch-video— transcript / summary output often flows throughpastefor platform-specific reformatting
Notes on quality
- Secret detection runs first, always. Before any formatting or destination logic,
pastescans forsk-*,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[31metc.) 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
twitterorlinkedin. 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
Sicherheitsprüfung
Quellcode ansehen
coreyhaines31/makerskills
Mehr aus dieser Quelle
Give agents clean web context
Search and extract the public web as Markdown or structured JSON through one API or hosted MCP server.
Verwandte Skills
Mehr aus dieser Quelle: coreyhaines31/makerskills
So verwenden Sie diesen Skill
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.
Keine Konfiguration erforderlich. Ihr KI-Agent (Claude Code, Cursor, Windsurf usw.) erkennt installierte Skills automatisch und nutzt sie als Kontext bei der Code-Generierung.
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.