Works with the AI Agents you already use
Spectre SkillGuard
Scan AI agent skills before installing or using them.
$15
Spectre SkillGuard
Example session with this skill installed
Scan the skill folder at ./tools/github-optimizer and tell me if it is safe to use.
- Read your context and instructions
- Compiled the spectre skillguard
I have completed the security audit of the github-optimizer skill located at ./tools/github-optimizer. The scan included a pattern pass for injection risks and a manual review of all command logic.
Security Report: github-optimizer
Verdict: ⚠️ Review before installing
Security Score: 69/100
Inventory Summary
- Entry Point:
SKILL.md(4.2 KB) - Scripts:
src/main.py,scripts/setup.sh - MCP Config: None detected
- Not checked:
bin/helper.exe(Binary file, skipped)
Findings
| ID | Severity | Location | Description |
|---|---|---|---|
| P2.5 | High | src/main.py:42 | os.system call uses unvalidated user input, leading to potential shell injection. |
| C3 | Medium | SKILL.md:12 | Skill requests gh auth but logic in main.py attempts to read ~/.ssh/id_rsa. |
| P4.1 | Low | scripts/setup.sh:5 | Hardcoded URL http://temp-binaries.xyz/dl found in download script. |
| P1.2 | Info | SKILL.md:88 | Instructions contain "ignore previous instructions" as a test case string. |
Verification of High Severity Finding
src/main.py:42:os.system("git commit -m " + user_input_msg)
Risk: An attacker could provide a message like"; rm -rf /; #"to execute arbitrary commands.
Next steps
- Sanitize the input in
src/main.pyusingsubprocess.runwith an argument list instead ofos.system. - Remove the unauthorized access attempt to the SSH directory.
- Replace the unverified HTTP download link with a pinned hash or official source.
Connects securely to your tools. The creator never sees your data.
About this skill
AI agent skills and MCP servers run with your full system permissions, creating a massive attack surface for prompt injection and data exfiltration. Spectre SkillGuard is a read-only security auditor that scans untrusted skills and plugins to determine if they are safe to install. It treats all target files as untrusted data, ensuring that malicious instructions within the code cannot hijack the auditing process itself.
What it does
- Multi-vector Scanning identifies prompt injection, secret exposure, dangerous shell execution, and supply-chain risks.
- Contextual Review analyzes whether a skill's promised behavior matches its actual internal logic and command execution.
- Static Pattern Analysis runs a specialized scanner to detect known malicious patterns, hidden files, and suspicious network calls.
- Risk Scoring calculates a 0-100 security score based on finding severity to provide a clear install/no-install verdict.
- Isolation Auditing detects symlink attacks and persistence mechanisms designed to break out of restricted environments.
How it works
- Target Selection provides a folder path, GitHub URL, or
.mcp.jsonfile for the agent to analyze. - Inventory & Pattern Pass performs a file-level inventory and executes a read-only pattern scan to flag immediate red flags.
- Manual Verification re-opens every cited finding to quote the exact line of code and map it to a specific security rule.
- Final Reporting delivers a structured security verdict (OK, Review, or Do Not Install) with a score and a list of specific findings.
Frameworks & tools
The skill uses Git for remote cloning, Perl for pattern scanning, and standard Unix-style file inspection tools. It analyzes SKILL.md files, MCP configurations, and scripts written in Bash, Python, JavaScript, and TypeScript.
Why this beats prompting it yourself
Generic prompts often fall victim to the very injections they are supposed to detect, as the agent may follow instructions found inside the malicious file. This skill enforces a strict read-only execution policy and a standardized scoring rubric that prevents the auditor from being compromised by the target code.
Use cases
- Third-party Auditing before installing a community-made Claude plugin or MCP server.
- CI/CD Safety Checks to scan changed skill files in a repository before merging them into a production agent.
- Secret Detection to ensure internal API tokens or SSH keys aren't accidentally bundled in a skill's supporting files.
- Functionality Verification to ensure a skill isn't making unauthorized
curlrequests to external domains.
Known limitations
Does not perform dynamic analysis or runtime monitoring of active agents. Requires Perl and Git to be available on the host system for full automated scanning and remote repo retrieval.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Security.
Requires Perl for automated pattern scanning. Does not perform dynamic sandbox execution. Cannot scan binary-only skills without source code.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 12 days ago
- Software
Python
Git