@plurnk/plurnk-execs-common

The universal subprocess executor for plurnk-service's exec scheme: one package covering the shell + node + python floor and whichever host interpreters (perl/ruby/php/lua/awk/bc/…) are present. Install it once and the model gets every subprocess runtime the host can serve — node always, the rest detected.

A @plurnk/plurnk-execs-* sibling built on the plurnk-execs framework. Supersedes the former -sh, -node, -python packages (folded in here).

How it works

The manifest claims the subprocess runtime tags. The framework's probe() (per-tag) lights up node unconditionally (the daemon is node) and detects the rest via a cheap command -v, so the consumer offers the model exactly the platform's runtimes. No per-language packages — one executor adapts to the host.

| Tag | Binary | Command via | |---|---|---| | sh 🐚 / bash 🐚 | sh / bash | -c <command> | | node ⬢ | node | -e <command> (always available) | | python / python3 🐍 | python3 | -c <command> | | perl 🐪 / ruby 💎 / lua 🌙 | perl / ruby / lua | -e <command> | | php 🐘 | php | -r <command> | | deno 🦕 | deno | eval <command> | | bun 🥟 | bun | -e <command> | | tcl 🪶 | tclsh | stdin | | bc 🧮 | bc | stdin (e.g. 6 * 7) | | awk 🪄 | awk | program arg, empty stdin (BEGIN { … }) |

All run arbitrary code → effect: host → proposal-gated (not auto-run). For the snappy auto-run tier, see the pure in-process evaluators (calc/jq/wasm, planned). Input-processing transforms (sed, input-driven awk) await the EXEC input-channel and aren't claimed here.

Configuration

.env.example documents per-tag kill-switches: PLURNK_EXECS_<TAG>=0 disables a tag even when installed (drops it from the available list with a "disabled" 501 reason).

Tests

test:lint, test:unit. Live-eval tests auto-skip where the interpreter is absent.