agentsclimarketplace

Cast record

Skill Paldom/terminaltor/skills/cast-record

Agent Skills for recording, sanitizing, and rendering terminal session demos - capture real command walkthroughs with agents, redact sensitive output, and render polished casts for READMEs and docs.

Install
npx -y skills add Paldom/terminaltor --skill cast-record

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 15 days oldThe repository was created 15 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Records real terminal sessions to asciicast .cast files with asciinema 3.x - interactive or headless/scripted capture with clean-environment hygiene so machine details and secrets stay out. Use when the user asks to record, capture, or rec a terminal session, command run, or shell activity. Not for authoring scripted demo tapes/gifs, redacting, or rendering casts.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

4.3 KB, as published. Nobody here has run it

cast-record

Capture a real terminal session — authentic output, colors, and timing — into an asciicast .cast file (newline-delimited JSON) with asciinema 3.x. The cast is the text-native source of truth: replayable in the terminal, diffable, and renderable later. Recording is local-first: nothing uploads unless you explicitly run asciinema upload.

When NOT to use

  • A polished, scripted README demo with simulated typing → tape-demo (nothing real needs capturing).
  • Turning an existing .cast into GIF/MP4 → cast-render.
  • Removing sensitive data from a recording → cast-redact.
  • Hosting, streaming, or embedding the asciinema web player → out of scope.

Workflow

  1. Prepare a leak-resistant session (prevention reduces risk; redaction still gates publishing):
    • Work in a fresh temp directory; consider HOME=$(mktemp -d) for demos that print paths, or a stripped environment (env -i HOME=$TMP PATH=/usr/bin:/bin TERM=xterm-256color bash) — inherited variables like SSH_AUTH_SOCK or cloud creds otherwise remain reachable by whatever you record.
    • Fake credentials only; set a generic prompt (PS1='$ ') if your prompt shows user@host.
    • The -c command string and captured env are stored in the cast header — never put a secret in the command line.
  2. Record.
    • Scripted/agent-driven (reproducible, no TTY needed):
      asciinema rec --headless --window-size 100x30 -i 2 \
        -c ./demo.sh --overwrite demo.cast
      
    • Interactive (you drive): asciinema rec demo.cast — end with ctrl+d or exit; ctrl+\ pauses/resumes capture mid-session.
    • Flags that matter: --window-size COLSxROWS fixes dimensions (3.x replaced the old --cols/--rows); -i/--idle-time-limit 2 caps dead air; -t "Title" labels the recording.
    • Treat the recorded command like any executable code: review scripts before recording them, and don't record with production credentials in the environment.
  3. Verify. head -1 demo.cast must be a JSON header ({"version":3,"term":{"cols":100,"rows":30,...}}). Replay with asciinema play demo.cast to check content and pacing.
  4. Pick the format. 3.x writes asciicast v3 by default (relative timestamps, easiest to edit). For tools that only read v2: asciinema rec -f asciicast-v2 … or later asciinema convert -f asciicast-v2 demo.cast demo-v2.cast. asciinema convert demo.cast demo.txt exports plain text for docs.
  5. Quarantine, then hand off. Keep the raw cast out of git (e.g. name it *.raw.cast and gitignore that pattern) until it has been secret-scanned — run cast-redact before publishing, then cast-render for GIF/MP4.

Output spec

  • A .cast file with a valid JSON header line and NDJSON events, at the agreed dimensions, idle-capped, replayable with asciinema play.
  • No upload happened; raw cast not committed until scanned/redacted.

Gotchas

  • -I/--capture-input records every keystroke, including passwords typed with echo off. It is off by default; enable only when keystrokes are the point, and say so.
  • -i/--idle-time-limit stores a limit in metadata for players to honor — captured timing is unchanged (agg applies its own idle cap at render time).
  • Header leaks: command, title, and captured env (SHELL by default, more via --capture-env) all land in the header — mind what they contain.
  • Overwrite/append: asciinema refuses to overwrite existing files without --overwrite; -a/--append continues a previous recording instead.
  • No Windows support (GNU/Linux, macOS, FreeBSD). Verified against asciinema 3.1.0; 3.x line current as of mid-2026 (3.2.1).
  • Everything printed to the terminal is in the cast — including text later cleared or backspaced over. Assume the file contains more than the final screen showed.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.