Security audit
Security checks for vulnerabilities and agentic risk
This is a straightforward Pruna image-upscaling guide, with disclosed image upload/API use and some unpinned related-skill install commands users should verify before running.
Before installing or following the setup commands, verify the PrunaAI source and consider pinning trusted versions of the related skills. Use this only with images and API credentials you are comfortable sending to Pruna's hosted service, and avoid enabling `disable_safety_checker` unless you understand the policy and output implications.
The skill instructs users to execute npx skills add ... commands without pinning an exact immutable version of the installer/tooling, which can cause future installs to pull changed code. If the skills package or resolution path is compromised or updated maliciously, an agent or operator could install and run unintended code.
This line references npx skills without an explicit pinned version, creating a supply-chain risk because the executed package can change over time. In agent workflows, this is especially risky because install commands may be followed automatically or with limited review.
The unpinned npx skills invocation can resolve to different code at different times, enabling package substitution or compromised upstream release abuse. Because this skill encourages prerequisite installation before use, it increases the chance that operators will execute the command directly.
Using npx skills without version pinning introduces a real supply-chain exposure: future executions may fetch modified code. The risk is not merely theoretical because this is presented as a setup step that users are expected to run.
This alternate-skill recommendation again includes an unpinned npx skills install command, carrying the same remote code execution/supply-chain risk as the prerequisite lines. Repetition across the file broadens the exposure surface because multiple fallback paths prompt execution of the same unpinned tool.
The file recommends installing another skill through an unpinned npx skills command, which can pull different code over time. In an agent skill context, these commands may be trusted as canonical setup instructions, increasing operational risk.
An unversioned npx skills command is a true supply-chain weakness because it delegates trust to the current state of a remote package at runtime. If the upstream package or namespace is hijacked, this could lead to arbitrary code execution on the user's machine.
This line repeats the same unpinned package execution pattern, which is dangerous because npx may download and execute remote package code. The context does not mitigate the risk; it normalizes running unaudited install commands as part of routine use.
The fallback install guidance uses npx skills without pinning, which is a valid supply-chain finding. If exploited through a compromised upstream package or dependency confusion scenario, it could execute attacker-controlled code during installation.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
curl -X POST "https://api.pruna.ai/v1/files" \
-H "apikey: ${PRUNA_API_KEY}" \
-F "content=@/path/to/source.png"
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
curl -X POST "https://api.pruna.ai/v1/files" \
-H "apikey: ${PRUNA_API_KEY}" \
-F "content=@/path/to/source.png"
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
curl -X POST "https://api.pruna.ai/v1/files" \
-H "apikey: ${PRUNA_API_KEY}" \
-F "content=@/path/to/source.png"
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
curl -X POST "https://api.pruna.ai/v1/files" \
-H "apikey: ${PRUNA_API_KEY}" \
-F "content=@/path/to/source.png"
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
curl -X POST "https://api.pruna.ai/v1/files" \
-H "apikey: ${PRUNA_API_KEY}" \
-F "content=@/path/to/source.png"
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
curl -X POST "https://api.pruna.ai/v1/files" \
-H "apikey: ${PRUNA_API_KEY}" \
-F "content=@/path/to/source.png"
The follow-on skill recommendation again instructs the use of unpinned npx skills, preserving the same package-fetch execution risk. Because it appears in 'next steps', users may execute it later with less scrutiny, extending exposure beyond initial setup.
This line contains another unpinned npx skills invocation, making it a true positive for supply-chain risk. The danger is amplified by the instructional nature of the document, which encourages execution rather than treating the command as illustrative only.
No suspicious patterns detected.