# Security Monitor

> >

- **Type:** Skill
- **Install:** `agentstack add skill-rationaleyes-claude-skills-security-guide-security-monitor`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [RationalEyes](https://agentstack.voostack.com/s/rationaleyes)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [RationalEyes](https://github.com/RationalEyes)
- **Source:** https://github.com/RationalEyes/claude-skills-security-guide/tree/main/skills/security-monitor

## Install

```sh
agentstack add skill-rationaleyes-claude-skills-security-guide-security-monitor
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Security Monitor Skill

This skill scans all installed Claude Skills for indicators of compromise and
injection attack patterns. It provides automated static analysis to detect
malicious patterns before untrusted skills are executed.

## What This Skill Does

When invoked, this skill runs `scripts/scan_skills.py` against all skill
directories discoverable from standard install locations:

- `~/.claude/skills/` (personal skills)
- `.claude/skills/` (project skills in current working directory)

The scanner checks each `SKILL.md` and any associated scripts for:

1. **External URL references** — C2 callbacks, data exfiltration endpoints
2. **Credential access patterns** — environment variable reads of sensitive keys
3. **Persistence indicators** — crontab edits, shell RC file writes, skill replication
4. **Trigger hijacking patterns** — overly broad or authority-impersonating descriptions
5. **Suspicious tool grants** — unrestricted Bash in `allowed-tools` without audit trail
6. **Hidden content** — HTML comments, zero-width characters, base64-encoded blobs

## Invocation

To run a full security audit of all installed skills:

```
Please run the security monitor skill to audit my installed skills.
```

The skill will execute the scanner and return a structured report. Each skill
receives a **risk rating**: `LOW`, `MEDIUM`, `HIGH`, or `CRITICAL`.

## Scanning a Specific Directory

To scan a skill pack or directory before installing:

```
Please scan the skills in ./downloaded-skills/ for security issues.
```

## Interpreting Results

- **LOW**: No significant indicators found. Normal use.
- **MEDIUM**: Patterns present that warrant review (e.g., external URLs in a skill
  that shouldn't need network access).
- **HIGH**: Multiple indicators or a single high-confidence indicator. Do not execute
  until manually reviewed.
- **CRITICAL**: Strong evidence of malicious intent (credential exfiltration,
  persistence, C2 patterns). Quarantine immediately.

## Important Notes

- This skill itself requires Bash access to run the Python scanner.
- The scanner is read-only and does not modify any skill files.
- False positives are possible — a legitimate skill may reference URLs for
  documentation purposes. Use the report as a starting point for human review,
  not as a definitive verdict.
- This skill does NOT prevent execution of other skills — it only reports findings.
  Enforcement requires a human decision to remove or quarantine flagged skills.

## Running the Scanner Manually

```bash
python3 ~/.claude/skills/security-monitor/scripts/scan_skills.py \
    --paths ~/.claude/skills .claude/skills \
    --report-format text
```

---

*Part of the Claude Skills Defense Suite.*

## Execution

Run the security scanner now against all discoverable skill locations:

```bash
python3 "$(dirname "$0")/scripts/scan_skills.py" \
    --paths ~/.claude/skills ./.claude/skills \
    --report-format text
```

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [RationalEyes](https://github.com/RationalEyes)
- **Source:** [RationalEyes/claude-skills-security-guide](https://github.com/RationalEyes/claude-skills-security-guide)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-rationaleyes-claude-skills-security-guide-security-monitor
- Seller: https://agentstack.voostack.com/s/rationaleyes
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.