Is awscli safe?

https://clawhub.ai/openclaw/awscli

72
CAUTION

The 'awscli' skill failed to install (timeout) and delivered no functional content — the SKILL.md is empty and the lock.json references a different skill name ('academic-research-hub'). During the failed install, the process accessed sensitive files including .aws/credentials and .env, though no network exfiltration was detected. The combination of credential access, install failure, and name mismatch warrants caution.

Category Scores

Prompt Injection 90/100 · 30%
Data Exfiltration 55/100 · 25%
Code Execution 70/100 · 20%
Clone Behavior 65/100 · 10%
Canary Integrity 100/100 · 10%
Behavioral Reasoning 60/100 · 5%

Findings (8)

HIGH AWS credentials accessed during install -30

The install process opened and read /home/oc-exec/.aws/credentials. While this may be part of the OpenClaw platform's config resolution rather than the skill itself, accessing cloud credentials during skill installation — before the user has explicitly authorized AWS access — is a significant trust concern.

MEDIUM .env file accessed during install -15

The installer read /home/oc-exec/.env which may contain secrets, API keys, and other sensitive environment variables.

MEDIUM JIT code compilation during install -15

The install process created 14+ compiled JavaScript files in /tmp/jiti/, indicating runtime code compilation and execution. This is likely platform behavior (jiti transpiler) but means untrusted code paths are executed during install.

MEDIUM Install timeout failure -20

The skill install failed with a timeout error, meaning the skill is non-functional. The installer hung long enough to trigger the timeout, which could indicate a heavy dependency resolution, a network request that was blocked, or a deliberate stalling technique.

MEDIUM Skill name mismatch in lock.json -20

The lock.json references 'academic-research-hub' as the installed skill rather than 'awscli'. This naming discrepancy could indicate a misconfigured package, a dependency pulled in unexpectedly, or a bait-and-switch where a different skill is actually being installed.

LOW Gateway lock file created -5

A lock file was created at /tmp/openclaw-1000/gateway.e9191928.lock during install, suggesting the platform attempted to start a local gateway process. While likely platform behavior, it indicates background process management during install.

INFO No network egress detected 0

Despite accessing credentials and config files, no outbound network connections were observed. This is a positive indicator that no immediate data exfiltration occurred during the monitored install window.

INFO Empty SKILL.md — no prompt content delivered 0

The skill contains no SKILL.md file, meaning it provides no instructions to the LLM agent. This means there are no prompt injection vectors, but also means the skill is non-functional as a prompt-based skill.