Security audit
Security checks for vulnerabilities and agentic risk
The skill is a disclosed guide for using Pruna's video-replacement API, with expected uploads and API calls for that purpose.
Before installing, confirm you are comfortable sending the source video and reference images to Pruna and using a PRUNA_API_KEY. For controlled environments, pin or review the referenced PrunaAI/pruna-skills installs instead of relying on mutable npx references, and avoid using safety-disabling options unless your policy explicitly allows them.
The skill instructs users to execute npx skills add ... without pinning the fetched package/skill to an immutable version or digest. That creates a supply-chain risk: if the upstream package/tag is changed or compromised, future installs may pull attacker-controlled code or altered skill content.
This line references an external skill via npx skills without an immutable version pin. Because skill installation executes/fetches remote content, a later upstream change could silently alter behavior or introduce malicious code into agent environments.
The pruna-api prerequisite is installed through an unpinned npx skills reference, enabling supply-chain substitution if the remote package or tag changes. In an agent workflow, this is particularly risky because the fetched skill can influence subsequent networked actions and credential handling.
The full-suite install command uses an unpinned remote skill bundle, increasing blast radius because a single floating reference can change multiple transitive skills at once. If the package is compromised upstream, users may install a broad set of malicious or unsafe instructions.
This alternate-skill recommendation still tells users to install via unpinned npx skills, preserving the same supply-chain exposure. Even if the primary skill is safe, redirect paths can become the compromise vector.
The p-video-animate recommendation is an unpinned remote dependency, allowing silent drift or malicious replacement over time. Users following the documentation may install different code than was originally reviewed.
This line recommends p-video-avatar through an unpinned install command, which is a classic documentation-level supply-chain weakness. Attackers often exploit exactly these copy-paste install paths because users trust official setup instructions.
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-video.mp4"
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-video.mp4"
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-video.mp4"
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-video.mp4"
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-video.mp4"
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-video.mp4"
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-video.mp4"
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-video.mp4"
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-video.mp4"
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-video.mp4"
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-video.mp4"
The follow-on p-video-animate install remains unpinned, so users can be exposed to remote code/content changes after completing the main workflow. This extends supply-chain risk beyond initial setup into subsequent operational steps.
This line documents installation of p-image without an immutable pin, making future installs non-reproducible and vulnerable to upstream compromise. In agent ecosystems, such drift can alter prompts, network behavior, or safety controls.
The p-image-edit install command is unpinned, which means the referenced remote skill can change after review. That is a real supply-chain vulnerability because users are encouraged to execute a network install based on mutable documentation.
The video-editing follow-on install also lacks version pinning, leaving users exposed to malicious or unintended upstream changes. Since this is presented as normal workflow guidance, it could be widely copy-pasted into trusted environments.
No suspicious patterns detected.