skills.rest
Vimurai
VimuraiCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

ai-analyze

Analyzes shell commands for risk and returns PASS, WARN, or BLOCK verdicts before execution.

Updated Feb 26, 2026
Search Tags:#mcp#risk assessment#shell#pre-flight check#destructive commands#security gate#command analysis
One-click install
npx skills add https://github.com/Vimurai/ai-os --skill ai-analyze-vimurai
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ai-analyze
Source: https://github.com/Vimurai/ai-os/tree/main/.claude/skills/ai-analyze
Command: npx skills add https://github.com/Vimurai/ai-os --skill ai-analyze-vimurai

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Running destructive or unfamiliar shell commands without a safety check can cause data loss or violate role boundaries in multi-agent workflows. This Skill pre-flights any shell command through the safe-exec-mcp analyze_command tool so you get a risk verdict before anything executes. ## Core Features & Use Cases - Risk Classification: Wraps the safe-exec-mcp analyze_command MCP tool to classify commands as PASS, WARN, or BLOCK. - Sovereignty Enforcement: Detects role violations (e.g., an Architect attempting an Engineer-only operation) via the injected AI_OS_CALLER_ROLE. - Use Case: Before running git reset --hard or an unfamiliar generated command, invoke the skill to get a verdict; on BLOCK it refuses and proposes a safe alternative. ## Quick Start Ask the assistant to analyze the shell command you are about to run, for example by saying "analyze this command: git push --force".

Frequently Asked Questions about ai-analyze

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I check if a shell command is safe before running it?▼

Invoke the skill with the exact command string and it calls the safe-exec-mcp analyze_command tool, which returns a PASS, WARN, or BLOCK verdict. On PASS you proceed; on WARN you confirm intent first; on BLOCK the command is refused.

What does the analyze_command MCP tool return?▼

It returns one of three verdicts: PASS for safe commands, WARN for risky but allowed commands, and BLOCK for destructive or sovereignty-violating commands. A [SOVEREIGNTY_BLOCK] indicates a role boundary violation that must be handed to the correct role.

Does the skill execute the command after analyzing it?▼

No, the skill is read-only over the command string and never executes it. You run the command yourself only after receiving a PASS verdict or confirming a WARN.

Should I pass caller_role manually to analyze_command?▼

No, the bootloader injects the tamper-resistant AI_OS_CALLER_ROLE automatically, and the MCP server prioritizes it over any manually passed argument. Passing it manually is unnecessary and ignored.

When should I use safe-exec --check instead of this skill?▼

Use safe-exec --check when you need a full local pre-execution gate that exits with code 2 on BLOCK, such as in scripts or CI pipelines. This skill is the interactive, single-command conversational form.

Related Skills

Explore More Software Engineering Skills→
mrmtsuruyamrmtsuruya

himalaya

Manage email via terminal using Himalaya CLI with IMAP and SMTP backends.

Community
Intermediate
noname300989noname300989

terminal-env

Manage sandboxed terminal sessions with tmux and timestamped command logging.

Community
Advanced
mrkhachaturovmrkhachaturov

careful

Warn before destructive Bash, Git, Kubernetes, and Docker commands with explicit override.

Community
Advanced
pantheon-orgpantheon-org
9

bash-script-validator

Validate bash and shell scripts for syntax errors and security vulnerabilities.

Official
Intermediate