Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generate bash script templates with best practices including error handling, argument parsing, logging, and portability considerations.
.claude/skills/a5c-ai-bash-script-template/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -6% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 15% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 27% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 24% | 0% |
Generate bash script templates with best practices.
Invoke this skill when you need to:
| Parameter | Type | Required | Description | |-----------|------|----------|-------------| | scriptName | string | Yes | Script name | | description | string | Yes | Script description | | arguments | array | No | Script arguments | | functions | array | No | Functions to include |
bash#!/usr/bin/env bash # # <scriptName> - <description> # # Usage: <scriptName> [options] <arguments> # set -euo pipefail IFS=$'\n\t' # Script directory readonly SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" readonly SCRIPT_NAME="$(basename "${BASH_SOURCE[0]}")" # Colors (if terminal supports) if [[ -t 1 ]]; then readonly RED='\033[0;31m' readonly GREEN='\033[0;32m' readonly YELLOW='\033[0;33m' readonly BLUE='\033[0;34m' readonly NC='\033[0m' else readonly RED='' GREEN='' YELLOW='' BLUE='' NC='' fi # Logging functions log_info() { echo -e "${BLUE}[INFO]${NC} $*"; } log_success() { echo -e "${GREEN}[OK]${NC} $*"; } log_warn() { echo -e "${YELLOW}[WARN]${NC} $*" >&2; } log_error() { echo -e "${RED}[ERROR]${NC} $*" >&2; } die() { log_error "$*"; exit 1; } # Cleanup on exit cleanup() { # Add cleanup tasks here : } trap cleanup EXIT # Usage usage() { cat <<EOF Usage: ${SCRIPT_NAME} [options] <argument> <description> Options: -h, --help Show this help message -v, --verbose Enable verbose output -d, --dry-run Show what would be done -f, --force Force operation -o, --output Output file path Arguments: argument Required argument description Examples: ${SCRIPT_NAME} -v input.txt ${SCRIPT_NAME} --output result.txt input.txt EOF } # Parse arguments parse_args() { local OPTIND opt while getopts ":hvdfo:-:" opt; do case "${opt}" in h) usage; exit 0 ;; v) VERBOSE=true ;; d) DRY_RUN=true ;; f) FORCE=true ;; o) OUTPUT="${OPTARG}" ;; -) case "${OPTARG}" in help) usage; exit 0 ;; verbose) VERBOSE=true ;; dry-run) DRY_RUN=true ;; force) FORCE=true ;; output=*) OUTPUT="${OPTARG#*=}" ;; *) die "Unknown option: --${OPTARG}" ;; esac ;; :) die "Option -${OPTARG} requires an argument" ;; ?) die "Unknown option: -${OPTARG}" ;; esac done shift $((OPTIND - 1)) # Required arguments if [[ $# -lt 1 ]]; then usage die "Missing required argument" fi ARGUMENT="$1" } # Defaults VERBOSE=${VERBOSE:-false} DRY_RUN=${DRY_RUN:-false} FORCE=${FORCE:-false} OUTPUT=${OUTPUT:-} # Main function main() { parse_args "$@" if [[ "${VERBOSE}" == true ]]; then log_info "Verbose mode enabled" fi if [[ "${DRY_RUN}" == true ]]; then log_warn "Dry run mode - no changes will be made" fi log_info "Processing: ${ARGUMENT}" # Main logic here log_success "Done!" } main "$@"
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 21,900 | 10,447 | -52% | 1 | 1 | 0% | 4,822 | 3,315 | -31% | 0 | 0 | — |
case-02 | fail→pass | 18,610 | 13,586 | -27% | 1 | 1 | 0% | 4,303 | 4,040 | -6% | 0 | 0 | — |
case-03 | fail→fail | 18,166 | 13,117 | -28% | 1 | 1 | 0% | 4,076 | 4,097 | +1% | 0 | 0 | — |
case-04 | fail→fail | 15,173 | 26,082 | +72% | 1 | 1 | 0% | 3,481 | 4,647 | +33% | 0 | 0 | — |
case-05 | fail→fail | 9,630 | 7,851 | -18% | 1 | 1 | 0% | 2,060 | 2,706 | +31% | 0 | 0 | — |
case-06 | fail→fail | 10,463 | 9,858 | -6% | 1 | 1 | 0% | 2,156 | 3,064 | +42% | 0 | 0 | — |
case-07 | fail→fail | 12,981 | 9,794 | -25% | 1 | 1 | 0% | 2,831 | 3,083 | +9% | 0 | 0 | — |
case-08 | pass→fail | 13,853 | 14,201 | +3% | 1 | 1 | 0% | 2,712 | 4,120 | +52% | 0 | 0 | — |
case-09 | fail→pass | 15,745 | 12,426 | -21% | 1 | 1 | 0% | 3,367 | 3,864 | +15% | 0 | 0 | — |
case-10 | fail→pass | 12,222 | 11,239 | -8% | 1 | 1 | 0% | 2,533 | 3,228 | +27% | 0 | 0 | — |
case-11 | fail→fail | 16,734 | 10,127 | -39% | 1 | 1 | 0% | 3,428 | 3,269 | -5% | 0 | 0 | — |
case-12 | fail→fail | 14,625 | 15,629 | +7% | 1 | 1 | 0% | 2,975 | 4,521 | +52% | 0 | 0 | — |
case-13 | fail→fail | 12,991 | 11,206 | -14% | 1 | 1 | 0% | 2,722 | 3,563 | +31% | 0 | 0 | — |
case-14 | fail→fail | 12,642 | 8,123 | -36% | 1 | 1 | 0% | 2,671 | 2,812 | +5% | 0 | 0 | — |
case-15 | fail→pass | 12,484 | 10,785 | -14% | 1 | 1 | 0% | 2,711 | 3,353 | +24% | 0 | 0 | — |
case-16 | fail→fail | 13,074 | 12,872 | -2% | 1 | 1 | 0% | 2,703 | 3,911 | +45% | 0 | 0 | — |
case-17 | fail→fail | 15,348 | 10,526 | -31% | 1 | 1 | 0% | 3,311 | 3,298 | -0% | 0 | 0 | — |
case-18 | fail→fail | 12,331 | 8,919 | -28% | 1 | 1 | 0% | 2,635 | 3,035 | +15% | 0 | 0 | — |
case-19 | fail→fail | 12,279 | 10,159 | -17% | 1 | 1 | 0% | 2,451 | 3,278 | +34% | 0 | 0 | — |
case-20 | fail→fail | 15,836 | 10,141 | -36% | 1 | 1 | 0% | 3,002 | 3,319 | +11% | 0 | 0 | — |
case-21 | fail→fail | 12,219 | 10,301 | -16% | 1 | 1 | 0% | 2,729 | 3,311 | +21% | 0 | 0 | — |
case-22 | fail→fail | 8,855 | 9,368 | +6% | 1 | 1 | 0% | 1,734 | 3,104 | +79% | 0 | 0 | — |
case-23 | fail→fail | 14,550 | 10,700 | -26% | 1 | 1 | 0% | 3,069 | 3,264 | +6% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted. The headline lift of +17 percentage points is the difference between those two pass rates over the 23 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.