Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
{"version":"1.0.0","domain":"linux","triggers":"script safety, pre-flight checks, availability risk, script interaction, automation safety, service disruption, Linux operational safety","role":"implementation","scope":"implementation","output-format":"code","content-types":["code","guidance","do-dont","examples"]}
Safe Script Execution for Linux Infrastructure Automation
Infrastructure engineer performing pre-flight validation and interaction mapping for Linux automation scripts to identify availability risks, resource conflicts, and service disruption points before execution.
TL;DR Checklist
Verify script has set -euo pipefail and explicit error handling on every command
Map all file-system paths the script touches and confirm no data-loss operations without backup
Check every systemctl call — would restarting this service cause downtime for dependent services?
Verify no rm -rf, dd, truncate, or filesystem resize operations without explicit user confirmation
Confirm all network ports the script opens are within the approved range and documented
Validate that the script runs under a dedicated service account, not root, unless explicitly required
When to Use
Use this skill when:
Deploying automation scripts on production hosts — Any script that modifies system state, restarts services, or changes configurations
Reviewing third-party or vendor scripts — Evaluating scripts from external sources before running them on your infrastructure
Implementing CI/CD pipeline steps that touch production systems — Ensuring automated deployment steps don't disrupt running services
Auditing existing runbooks and playbooks — Identifying risky operations in documented procedures
Preparing to run migration or upgrade scripts — Systems undergoing version changes where downtime or data loss is unacceptable
Validating infrastructure-as-code changes before apply — Terraform, Ansible, or Salt states that modify live systems
When NOT to Use
Avoid this skill for:
Local development or sandbox environments — Where no availability risk exists, standard testing is sufficient
— , , , do not modify system state
Read-only inspection commands
systemctl status
journalctl
df
free
One-time setup on disposable infrastructure — Where the entire VM/container is ephemeral and can be discarded
Network packet capture or passive monitoring — Tools like tcpdump or ss that observe without modifying
Use linux-service-integrity-operations when the script is ready to execute and you need safe restart patterns and health monitoring during the change.
Core Workflow
1. Parse Script for Dangerous Operations
Identify all operations that modify system state. Build an inventory of every file, service, and network endpoint the script touches.
#!/usr/bin/env bash# preflight_audit.sh — Analyze a deployment script for availability risks# Usage: ./preflight_audit.sh <script.sh>set -euo pipefail
readonly DANGEROUS_PATTERNS=(
'systemctl\s+(restart|stop|reload|kill)''rm\s+(-rf|--no-preserve-root)''dd\s+of=''truncate\s+-s\s+0''mkfs\.''lvremove|lvreduce|vgremove''iptables|nftables\s+-F''modprobe|insmod|rmmod''fallocate|truncate\s+-s''curl|wget.*\|.*bash|sh\s*-''chattr\s+-i''umount''swapoff'
)
readonly DANGEROUS_PATTERNS_NAMES=(
'service_restart''destructive_rm''disk_write''file_truncate''filesystem_create''lv_destruction''firewall_flush''kernel_module''disk_allocation''pipe_risk''attr_remove''unmount''swap_disable'
)
audit_script() {
local script_file="$1"local risk_level="LOW"local findings=()
local service_ops=()
local file_ops=()
local resource_risks=()
if [[ ! -f "$script_file" ]]; thenecho"ERROR: Script not found: $script_file" >&2
return 1
fiecho"=== Pre-flight Audit: $(basename "$script_file") ==="echo"File: $script_file"echo"Size: $(wc -c < "$script_file") bytes"echo""# Check for safety patternsif ! grep -q 'set -euo pipefail'"$script_file"; then
findings+=("CRITICAL: Missing 'set -euo pipefail' — unhandled errors will cascade")
risk_level="CRITICAL"fiif ! grep -qE '^\s*(set|trap)'"$script_file"; then
findings+=("WARNING: No error trapping or set flags — failures may go undetected")
fi# Check dangerous operationsfor i in"${!DANGEROUS_PATTERNS[@]}"; doif grep -qE "${DANGEROUS_PATTERNS[$i]}""$script_file"; thenlocal matches
matches=$(grep -cE "${DANGEROUS_PATTERNS[$i]}""$script_file")
findings+=("HIGH: ${DANGEROUS_PATTERNS_NAMES[$i]} detected — ${matches} occurrence(s)")
risk_level="HIGH"fidone# Check for hard-coded credentialsif grep -qiE '(password|secret|key|token)\s*=\s*["\x27]'"$script_file"; then
findings+=("CRITICAL: Hard-coded credentials detected in script")
risk_level="CRITICAL"fi# Map systemctl operationswhile IFS= read -r line; do
service_ops+=("$line")
done < <(grep -nE 'systemctl\s+(restart|stop|reload|kill|disable|enable)'"$script_file" 2>/dev/null || true)
if [[ ${#service_ops[@]} -gt 0 ]]; thenecho""echo"=== Service Impact Analysis ==="for op in"${service_ops[@]}"; dolocal svc_name
svc_name=$(echo"$op" | grep -oE '[a-zA-Z0-9_-]+\.service' || echo"unknown service")
echo" [${op}] → Service: $svc_name"donefiecho""echo"=== Risk Assessment ==="echo"Overall risk: $risk_level"echo"Findings: ${#findings[@]}"for finding in"${findings[@]}"; doecho" • $finding"doneecho""echo"=== Recommendations ==="if [[ "$risk_level" == "CRITICAL" ]]; thenecho" • DO NOT run this script in production without remediation"echo" • Address all CRITICAL findings before proceeding"echo" • Obtain approval from on-call engineer"elif [[ "$risk_level" == "HIGH" ]]; thenecho" • Review all HIGH findings before execution"echo" • Schedule during maintenance window"echo" • Ensure rollback procedure is ready"elseecho" • Standard change process applies"echo" • Ensure monitoring is active during execution"fi
}
audit_script "${1:?Usage: $0 <script.sh>}"
Checkpoint: All dangerous operations are identified and catalogued. No CRITICAL findings remain unresolved.
2. Map Service Dependencies
Determine what other services depend on any service this script will restart or modify. Use systemctl list-dependencies to map the full dependency tree.
#!/usr/bin/env bash# service_dependency_map.sh — Map full dependency impact before restarting a service# Usage: ./service_dependency_map.sh <service.service>set -euo pipefail
map_dependencies() {
local target_service="$1"if ! systemctl list-unit-files "${target_service}" &>/dev/null; thenecho"ERROR: Service '${target_service}' not found" >&2
return 1
fiecho"=== Dependency Impact Analysis: ${target_service} ==="echo""# Upstream dependencies — what must be running for this service to workecho"--- Upstream Dependencies (Required Before Restart) ---"
systemctl list-dependencies --reverse "${target_service}" --no-pager 2>/dev/null | \
sed 's/├── / /; s/└── / /; s/│ / /' | \
grep -v 'list-dependencies' || echo" (none)"echo""# Downstream dependents — what breaks if this service restartsecho"--- Downstream Dependents (At Risk During Restart) ---"
systemctl list-dependencies "${target_service}" --no-pager 2>/dev/null | \
sed 's/├── / /; s/└── / /; s/│ / /' | \
grep -v 'list-dependencies' || echo" (none)"echo""# Check for network ports that would be briefly unavailableecho"--- Network Port Impact ---"local socket_unit
socket_unit=$(systemctl cat"${target_service}" 2>/dev/null | grep -E '^ListenStream|^ListenDatagram' || true)
if [[ -n "$socket_unit" ]]; thenecho" Socket activations detected:"echo"$socket_unit" | while IFS= read -r line; doecho" $line"donefi# Check if this is a core infrastructure servicelocal critical_services=(
"systemd-journald""systemd-logind""dbus""NetworkManager""sshd""cron""systemd-timesyncd""containerd""docker""polkit""udev""runit"
)
for crit in"${critical_services[@]}"; doif [[ "$target_service" == "${crit}.service" || "$target_service" == "${crit}.timer" ]]; thenecho""echo" ⚠ CRITICAL: '${target_service}' is a core infrastructure service"echo" ⚠ Restart may cause brief system-wide instability"echo" ⚠ Use 'systemctl reload' instead of 'restart' where possible"echo" ⚠ Ensure remote access is available via console/serial before proceeding"breakfidone# Check for OnFailure handlerslocal failure_handler
failure_handler=$(systemctl cat"${target_service}" 2>/dev/null | grep '^OnFailure=' || true)
if [[ -n "$failure_handler" ]]; thenecho""echo"--- Cascade Failure Risk ---"echo" OnFailure handler: $failure_handler"echo" If this service fails to restart, $failure_handler will be triggered"fi
}
map_dependencies "${1:?Usage: $0 <service.service>}"
Checkpoint: Full dependency tree is mapped. No core infrastructure services are targeted for restart without console access.
3. Verify Resource Headroom
Confirm the host has sufficient disk space, memory, and CPU headroom for the script's expected workload.
#!/usr/bin/env bash# resource_headroom.sh — Verify host has sufficient resources before running a script# Usage: ./resource_headroom.sh <script.sh>set -euo pipefail
check_resource_headroom() {
local script_file="$1"echo"=== Resource Headroom Check ==="echo""# Disk space analysisecho"--- Disk Space ---"local disk_usage
disk_usage=$(df -h / | awk 'NR==2 {print $5}' | tr -d '%')
local disk_free
disk_free=$(df -h / | awk 'NR==2 {print $4}')
if [[ "$disk_usage" -gt 90 ]]; thenecho" ⚠ CRITICAL: Root filesystem at ${disk_usage}% — script may fail"echo" ⚠ Free space: $disk_free"elif [[ "$disk_usage" -gt 80 ]]; thenecho" WARNING: Root filesystem at ${disk_usage}% — tight on space"echo" Free space: $disk_free"elseecho" OK: Root filesystem at ${disk_usage}% — sufficient space"echo" Free space: $disk_free"fi# Check if script creates/extends any fileslocal estimated_disk
estimated_disk=$(grep -oE '(truncate\s+-s\s+(\d+[kmgKMG])?)|(dd\s+.*count=\d+)|(mkfs|fdisk)'"$script_file" 2>/dev/null || true)
if [[ -n "$estimated_disk" ]]; thenecho" ⚠ Script contains disk-extending operations"echo" Operations: $estimated_disk"fiecho""# Memory analysisecho"--- Memory ---"local mem_total
mem_total=$(grep MemTotal /proc/meminfo | awk '{print int($2/1024)}')
local mem_available
mem_available=$(grep MemAvailable /proc/meminfo | awk '{print int($2/1024)}')
local mem_used_pct
mem_used_pct=$(awk "BEGIN {printf \"%d\", (1 - $mem_available / $mem_total) * 100}")
if [[ "$mem_used_pct" -gt 90 ]]; thenecho" ⚠ CRITICAL: Memory at ${mem_used_pct}% — ($mem_available MB available of ${mem_total} MB)"elif [[ "$mem_used_pct" -gt 80 ]]; thenecho" WARNING: Memory at ${mem_used_pct}% — ($mem_available MB available of ${mem_total} MB)"elseecho" OK: Memory at ${mem_used_pct}% — ($mem_available MB available of ${mem_total} MB)"fiecho""# CPU load checkecho"--- CPU Load ---"local load_avg
load_avg=$(cat /proc/loadavg | awk '{print $1}')
local cpu_count
cpu_count=$(nproc)
local load_ratio
load_ratio=$(awk "BEGIN {printf \"%.2f\", $load_avg / $cpu_count}")
if awk "BEGIN {exit !($load_ratio > 2.0)}"; thenecho" ⚠ WARNING: Load ratio ${load_ratio}x CPU count — consider delaying execution"elseecho" OK: Load ratio ${load_ratio}x CPU count (${load_avg} on ${cpu_count} CPUs)"fiecho""# File descriptor checkecho"--- File Descriptors ---"local fd_limit
fd_limit=$(ulimit -n)
local fd_used
fd_used=$(ls /proc/$$/fd 2>/dev/null | wc -l)
if [[ "$fd_limit" -lt 1024 ]]; thenecho" WARNING: File descriptor limit is low ($fd_limit)"elseecho" OK: FD limit $fd_limit, currently used $fd_used"fi
}
check_resource_headroom "${1:?Usage: $0 <script.sh>}"
Checkpoint: All resource metrics are within acceptable thresholds. No CRITICAL findings.
4. Generate Risk Assessment Report
Produce a structured summary of all findings, risk level, and required approvals before execution.
from dataclasses import dataclass, field
from enum import Enum
from typing importOptionalfrom datetime import datetime
classRiskLevel(Enum):
SAFE = "SAFE"
LOW = "LOW"
MEDIUM = "MEDIUM"
HIGH = "HIGH"
CRITICAL = "CRITICAL"@dataclassclassScriptFinding:
"""A single finding from pre-flight script analysis."""
severity: str# CRITICAL, HIGH, WARNING, INFO
category: str# error_handling, dangerous_op, service_impact, etc.
message: str
line_number: Optional[int] = None
remediation: str = ""@dataclassclassRiskAssessment:
"""Structured pre-flight risk assessment for a Linux automation script."""
script_path: str
risk_level: RiskLevel
findings: list = field(default_factory=list)
service_impacts: list = field(default_factory=list)
resource_risks: list = field(default_factory=list)
requires_approval: bool = False
recommended_window: str = ""defadd_finding(self, severity: str, category: str, message: str,
line_number: Optional[int] = None, remediation: str = ""):
self.findings.append(ScriptFinding(severity, category, message, line_number, remediation))
if severity == "CRITICAL":
self.risk_level = RiskLevel.CRITICAL
self.requires_approval = Trueelif severity == "HIGH"andself.risk_level notin (RiskLevel.CRITICAL,):
self.risk_level = RiskLevel.HIGH
self.requires_approval = Truedefto_report(self) -> str:
"""Generate a human-readable risk assessment report."""
lines = [
f"=== Pre-Flight Risk Assessment ===",
f"Script: {self.script_path}",
f"Risk Level: {self.risk_level.value}",
f"Timestamp: {datetime.now().isoformat()}",
f"Requires Approval: {'YES'if self.requires_approval else'No'}",
f"",
f"--- Findings ({len(self.findings)}) ---",
]
for f inself.findings:
prefix = "🔴"if f.severity == "CRITICAL"else"🟡"if f.severity == "HIGH"else"🔵"if f.severity == "WARNING"else"ℹ️"
line = f" {prefix} [{f.severity}] {f.category}: {f.message}"if f.line_number:
line += f" (line {f.line_number})"if f.remediation:
line += f"\n → {f.remediation}"
lines.append(line)
ifself.service_impacts:
lines.append("")
lines.append("--- Service Impact ---")
for imp inself.service_impacts:
lines.append(f" • {imp}")
lines.append("")
lines.append("--- Execution Recommendation ---")
ifself.risk_level == RiskLevel.CRITICAL:
lines.append(" DO NOT EXECUTE without remediation and approval.")
elifself.risk_level == RiskLevel.HIGH:
lines.append(" Schedule during maintenance window. Ensure rollback is ready.")
elifself.risk_level == RiskLevel.MEDIUM:
lines.append(" Standard change process applies. Monitor closely.")
else:
lines.append(" Low risk. Standard execution procedure applies.")
return"\n".join(lines)
defassess_script_risk(
script_path: str,
dangerous_ops: Optional[list[str]] = None) -> RiskAssessment:
"""Perform a pre-flight risk assessment on a shell script.
Analyzes the script for error handling, dangerous operations,
service dependencies, and resource consumption risks.
Args:
script_path: Path to the shell script to analyze
dangerous_ops: Optional list of additional dangerous patterns to check
Returns:
RiskAssessment with all findings and risk level
Raises:
FileNotFoundError: If script_path does not exist
PermissionError: If script_path is not readable
"""import os
import re
ifnot os.path.isfile(script_path):
raise FileNotFoundError(f"Script not found: {script_path}")
ifnot os.access(script_path, os.R_OK):
raise PermissionError(f"Script not readable: {script_path}")
assessment = RiskAssessment(script_path=script_path)
content = open(script_path).read()
lines = content.splitlines()
default_dangerous = [
("systemctl\\s+(restart|stop|reload|kill)", "service_restart"),
("rm\\s+(-rf|--no-preserve-root)", "destructive_rm"),
("(dd\\s+of=|truncate\\s+-s\\s+0)", "disk_write"),
("(iptables|nftables)\\s+-F", "firewall_flush"),
("(modprobe|insmod|rmmod)", "kernel_module"),
("curl.*\\|.*bash", "pipe_risk"),
]
patterns = dangerous_ops or default_dangerous
for i, line inenumerate(lines, start=1):
for pattern, category in patterns:
if re.search(pattern, line):
assessment.add_finding(
severity="HIGH"if"restart"in category else"CRITICAL",
category=category,
message=f"Dangerous operation in line {i}: {line.strip()}",
line_number=i,
remediation=f"Add --dry-run support and rollback before this operation"
)
# Check for credential patternsif re.search(r'(password|secret|token)\s*=\s*["\'][^\s"\']', line, re.IGNORECASE):
assessment.add_finding(
severity="CRITICAL",
category="credential_exposure",
message=f"Hard-coded credential detected",
line_number=i,
remediation="Move to environment variable or secrets manager"
)
# Check error handlingif"set -euo pipefail"notin content:
assessment.add_finding(
severity="CRITICAL",
category="error_handling",
message="Missing 'set -euo pipefail' — unhandled errors will cascade",
remediation="Add 'set -euo pipefail' as the second line of the script"
)
if"trap"notin content and"set -E"notin content:
assessment.add_finding(
severity="WARNING",
category="error_handling",
message="No error trapping configured",
remediation="Add 'trap cleanup EXIT ERR INT TERM' handler"
)
return assessment
Checkpoint: Risk assessment report is generated. All CRITICAL findings have remediation steps. Approval decision is documented.
Implementation Patterns
Pattern 1: Safe File Operation Wrapper
BAD — Unsafe file operations with no validation
#!/bin/bash# ❌ BAD: No error handling, no backup, no validationdeploy_config() {
cp /tmp/new-config.conf /etc/myapp/config.conf
systemctl restart myapp
echo"Deployed successfully"
}
# Problems:# - If cp fails, the old config is replaced with nothing (data loss)# - If restart fails, service is down with no recovery# - No backup of original config# - No post-deploy health check
GOOD — Safe file operations with validation and rollback
#!/usr/bin/env bash# ✅ GOOD: Safe deploy with backup, validation, and rollbackset -euo pipefail
readonly APP_NAME="myapp"readonly APP_DIR="/etc/${APP_NAME}"readonly BACKUP_DIR="${APP_DIR}/.backups"readonly CONFIG_FILE="${APP_DIR}/config.conf"deploy_config() {
local source_file="${1:?Usage: deploy_config <new-config.conf>}"local deploy_timestamp
deploy_timestamp="$(date +%Y%m%dT%H%M%S)"# Validate source exists and is readable
[[ -f "$source_file" ]] && [[ -r "$source_file" ]] || {
echo"ERROR: Source config not found or not readable: $source_file" >&2
return 1
}
# Validate destination directory exists
[[ -d "$APP_DIR" ]] || {
echo"ERROR: Application directory does not exist: $APP_DIR" >&2
return 1
}
# Backup current configmkdir -p "$BACKUP_DIR"local backup_file="${BACKUP_DIR}/config.conf.${deploy_timestamp}"cp -a "$CONFIG_FILE""$backup_file"echo"Backed up config to $backup_file"# Validate new config syntax before deployingifcommand -v "${APP_NAME}-ctl" &>/dev/null; thenif ! "${APP_NAME}-ctl" validate-config "$source_file" 2>/dev/null; thenecho"ERROR: New config failed syntax validation" >&2
echo"Restoring backup..."cp -a "$backup_file""$CONFIG_FILE"return 1
fifi# Deploy with correct ownershipcp"$source_file""${CONFIG_FILE}"chown root:root "$CONFIG_FILE"chmod 644 "$CONFIG_FILE"# Reload service (prefer reload over restart to minimize disruption)if systemctl reload "${APP_NAME}" &>/dev/null; thenecho"Service reloaded successfully"elif systemctl restart "${APP_NAME}" &>/dev/null; thenecho"Service restarted (reload unavailable)"elseecho"ERROR: Failed to reload/restart ${APP_NAME}" >&2
echo"Restoring backup..."cp -a "$backup_file""$CONFIG_FILE"echo"Config restored. Contact operations team."return 1
fi# Health check after deploysleep 2
if systemctl is-active --quiet "${APP_NAME}"; thenecho"Post-deploy health check passed"return 0
elseecho"ERROR: Health check failed after deploy" >&2
echo"Restoring backup and stopping service..."cp -a "$backup_file""$CONFIG_FILE"
systemctl restart "${APP_NAME}"return 1
fi
}
deploy_config "${1:?Usage: $0 <new-config.conf>}"
Pattern 2: Pre-flight Script Executor with Dry-Run
Bash — Safe execution framework with dry-run and rollback support
#!/usr/bin/env bash# safe_script_executor.sh — Execute scripts with pre-flight validation and rollback# Usage: ./safe_script_executor.sh [--dry-run] [--timeout 300] <script.sh>set -euo pipefail
DRY_RUN=false
TIMEOUT=300
SCRIPT_PATH=""# Parse argumentswhile [[ $# -gt 0 ]]; docase"$1"in
--dry-run) DRY_RUN=true; shift ;;
--timeout) TIMEOUT="$2"; shift 2 ;;
*) SCRIPT_PATH="$1"; shift ;;
esacdoneif [[ -z "$SCRIPT_PATH" ]]; thenecho"Usage: $0 [--dry-run] [--timeout N] <script.sh>" >&2
exit 1
fi# Create a rollback snapshot if this is a live run
ROLLBACK_SNAPSHOT=""create_rollback_snapshot() {
ROLLBACK_SNAPSHOT=$(mktemp -d "/tmp/safe-exec-rollback.XXXXXX")
# Snapshot mounted filesystems (only those we can safely snapshot)for mount_point in /etc /var/lib /opt; doif [[ -d "$mount_point" ]]; thenecho" Snapshotting: $mount_point"
rsync -a --delete "$mount_point/""${ROLLBACK_SNAPSHOT}${mount_point}/" 2>/dev/null || truefidone# Snapshot systemd service states
systemctl list-unit-files --state=enabled --no-pager 2>/dev/null | \
awk 'NR>1 {print $1}' > "${ROLLBACK_SNAPSHOT}/enabled-services.txt"echo"Rollback snapshot created: $ROLLBACK_SNAPSHOT"
}
rollback() {
if [[ -z "$ROLLBACK_SNAPSHOT" ]]; thenecho"WARNING: No rollback snapshot available" >&2
return 1
fiecho"=== Executing Rollback ==="# Restore configuration directoriesfor target in /etc /var/lib /opt; doif [[ -d "${ROLLBACK_SNAPSHOT}${target}" ]]; thenecho" Restoring: $target"
rsync -a "${ROLLBACK_SNAPSHOT}${target}/""$target/" 2>/dev/null || truefidoneecho"Rollback complete. Snapshot available at: $ROLLBACK_SNAPSHOT"echo"Review differences with: diff -r $target${ROLLBACK_SNAPSHOT}${target}"
}
trap rollback ERR
# Pre-flight checksecho"=== Pre-flight Checks ==="echo"Script: $SCRIPT_PATH"echo"Dry-run: $DRY_RUN"echo"Timeout: ${TIMEOUT}s"echo""# Check script exists and is readableif [[ ! -f "$SCRIPT_PATH" ]]; thenecho"ERROR: Script not found: $SCRIPT_PATH" >&2
exit 1
fiif [[ ! -r "$SCRIPT_PATH" ]]; thenecho"ERROR: Script not readable: $SCRIPT_PATH" >&2
exit 1
fi# Run the preflight audit if availableifcommand -v ./preflight_audit.sh &>/dev/null; thenecho"Running preflight audit..."
./preflight_audit.sh "$SCRIPT_PATH" || echo"WARNING: Preflight audit had findings — review above"echo""fi# Create rollback snapshot for live runsif [[ "$DRY_RUN" != "true" ]]; then
create_rollback_snapshot
echo""echo"WARNING: This is a LIVE run. A rollback snapshot has been created."echo"Press Ctrl+C within 5 seconds to cancel..."sleep 5
elseecho"DRY RUN mode — no changes will be made. No rollback needed."echo""fi# Execute the script with timeoutecho"=== Executing Script ==="if [[ "$DRY_RUN" == "true" ]]; then
bash -n "$SCRIPT_PATH" && echo"Syntax check passed" || {
echo"ERROR: Script has syntax errors" >&2
exit 1
}
echo"DRY RUN: Would execute: bash \"$SCRIPT_PATH\""else
bash "$SCRIPT_PATH" &
SCRIPT_PID=$!
# Monitor with timeout
(
sleep"$TIMEOUT"ifkill -0 "$SCRIPT_PID" 2>/dev/null; thenecho"ERROR: Script exceeded ${TIMEOUT}s timeout — killing" >&2
kill -TERM "$SCRIPT_PID" 2>/dev/null || truesleep 5
kill -9 "$SCRIPT_PID" 2>/dev/null || trueexit 1
fi
) &
MONITOR_PID=$!
wait"$SCRIPT_PID"
EXEC_EXIT=$?
# Kill the monitorkill"$MONITOR_PID" 2>/dev/null || truewait"$MONITOR_PID" 2>/dev/null || trueif [[ $EXEC_EXIT -ne 0 ]]; thenecho"ERROR: Script exited with code $EXEC_EXIT" >&2
echo"Rollback has been triggered automatically."exit 1
fiecho"Script completed successfully"fi# Cleanup rollback on successif [[ -n "$ROLLBACK_SNAPSHOT" && $EXEC_EXIT -eq 0 ]]; thenrm -rf "$ROLLBACK_SNAPSHOT"echo"Rollback snapshot cleaned up (deployment succeeded)"fiecho"=== Execution Complete ==="
Constraints
MUST DO
MUST require set -euo pipefail in all scripts before approving execution — scripts without it are unconditionally rejected
MUST map every service restart to its full dependency tree using systemctl list-dependencies before approving
MUST create a rollback snapshot or backup before any operation that modifies filesystem state on a running production host
MUST verify disk space and memory headroom before executing scripts that modify data or launch new processes
MUST use systemctl reload instead of systemctl restart wherever the service supports graceful reload
MUST run a post-execution health check after every service restart — never assume the service started correctly
MUST validate new configuration files with the application's native validation command before deploying them
MUST implement a timeout wrapper around every script execution to prevent hung processes from blocking the host
MUST NOT DO
MUST NOT run rm -rf or any destructive file operation on a production host without an explicit --yes flag and manual confirmation
MUST NOT restart systemd-journald, dbus, or udev on a running system — these can cause cascading failures
MUST NOT flush firewall rules (iptables -F or nft flush ruleset) without immediately restoring the allow rules
MUST NOT run scripts as root when a dedicated service account can perform the task
MUST NOT pipe remote data directly into a shell (curl URL | bash) — download, verify checksum, then execute
MUST NOT skip the pre-flight audit step regardless of script size or perceived trust level
MUST NOT hard-code credentials in scripts — use environment variables, systemd EnvironmentFile, or a secrets manager
Output Template
When applying this skill, produce:
Risk Assessment Summary — Risk level (SAFE/LOW/MEDIUM/HIGH/CRITICAL), total findings count, and whether approval is required
Dangerous Operations Inventory — Every file-system, service, and network operation the script performs, with line numbers and severity
Service Dependency Map — For every service the script restarts: upstream dependencies and downstream dependents
Resource Headroom Report — Current disk, memory, and CPU metrics with pass/fail against safety thresholds
Execution Decision — Clear recommendation: approved, approved-with-caveats, or blocked, with justification
Rollback Plan — Specific rollback steps and snapshot location if execution fails or causes unexpected disruption
Related Skills
Skill
Purpose
linux-service-integrity-operations
Execute changes without service interruption — use after pre-flight validation passes
linux-security
Validate that the script doesn't introduce security regressions (permissions, MAC policies)
linux-services
Understand systemd unit file structure and dependency ordering for accurate impact analysis
networking
Check network port allocations and firewall implications of the script's operations
Live References
Authoritative documentation links for this skill's domain. The model follows markdown links at load time to resolve external references and inline content.
