Security audit
Security checks for vulnerabilities and agentic risk
This is a coherent Pruna video-animation skill, with expected third-party media upload and some install/versioning cautions users should review.
Install only if you are comfortable sending the chosen image and video to Pruna for processing, and avoid confidential or unauthorized media. Review or pin the referenced Pruna companion skills before running the npx install commands, especially in shared or production agent environments.
The skill description activates on phrases like 'someone wants a photo to move like another video' and includes broad examples such as 'dance remixes' and 'performance variations from a template clip.' This gives general intent guidance but no explicit trigger boundaries or negative examples, which can cause unintended invocation for adjacent photo/video editing requests.
The skill instructs the agent to install and run npx skills packages without pinning an immutable version or commit. This creates a supply-chain risk: if the referenced package or tag is updated maliciously or unexpectedly, the agent may fetch and execute altered code during skill setup.
This line references an unpinned npx skills package install, allowing remote code or prompt content to change over time without review. In agent environments, such dynamic dependency resolution can lead to execution of attacker-controlled updates or prompt-injection content from a compromised upstream package.
An unpinned npx skills dependency is being recommended for installation, which exposes the workflow to package substitution or malicious upstream changes. Because this is an agent skill that influences future behavior, compromise could affect both code execution and model instructions.
The full-suite install command uses a floating reference, meaning the resolved content may differ across time and environments. This weakens reproducibility and opens a path for supply-chain compromise if the upstream package/tag is modified.
This install instruction points to an unpinned external skill package, which could later resolve to modified content without any change in the skill file. In a tool-using agent, that raises supply-chain and trust-boundary risks beyond a normal documentation issue.
The skill encourages installation of another unversioned remote package via npx skills, which is a true supply-chain exposure. If the package is compromised, an agent following this documentation may ingest malicious instructions or execute unintended code.
This line references a floating external dependency for skill installation. Unpinned dependencies are dangerous because they can change behavior later, including adding malicious logic or prompts that the agent would trust as prerequisites.
The install command for this related skill is not pinned, creating the same supply-chain risk as the other npx skills lines. Because the skill explicitly instructs agents to load prerequisites, this is more than an informational issue and directly affects runtime trust.
The skill instructs uploading user-provided image and video assets to a remote API but does not require an explicit user-facing disclosure or consent checkpoint about third-party transmission and privacy implications. For personal media, this can expose sensitive biometric, location, or proprietary content without adequately informing the user.
This command explicitly uploads a local video file to an external service. External transmission is expected for the skill's function, but it is still security-relevant because media may contain sensitive or regulated information and the skill does not enforce informed consent or minimization before transfer.
curl -X POST "https://api.pruna.ai/v1/files" \
-H "apikey: ${PRUNA_API_KEY}" \
-F "content=@/path/to/source-video.mp4"
This command explicitly uploads a local video file to an external service. External transmission is expected for the skill's function, but it is still security-relevant because media may contain sensitive or regulated information and the skill does not enforce informed consent or minimization before transfer.
curl -X POST "https://api.pruna.ai/v1/files" \
-H "apikey: ${PRUNA_API_KEY}" \
-F "content=@/path/to/source-video.mp4"
This command uploads a reference image to a third-party API, which may include facial or other sensitive visual data. In the context of an animation skill, such uploads are operationally necessary, but the lack of explicit privacy notice and consent makes the transmission security-relevant.
-H "apikey: ${PRUNA_API_KEY}"
-F "content=@/path/to/source-video.mp4"
curl -X POST "https://api.pruna.ai/v1/files"
-H "apikey: ${PRUNA_API_KEY}"
-F "content=@/path/to/reference-image.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/predictions' \
-H 'Content-Type: application/json' \
-H "apikey: ${PRUNA_API_KEY}" \
-H 'Model: p-video-animate' \
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/predictions' \
-H 'Content-Type: application/json' \
-H "apikey: ${PRUNA_API_KEY}" \
-H 'Model: p-video-animate' \
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/predictions' \
-H 'Content-Type: application/json' \
-H "apikey: ${PRUNA_API_KEY}" \
-H 'Model: p-video-animate' \
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/predictions' \
-H 'Content-Type: application/json' \
-H "apikey: ${PRUNA_API_KEY}" \
-H 'Model: p-video-animate' \
The request body sends URLs referencing uploaded user assets to the remote prediction endpoint, causing the service to process personal media off-box. This is contextually expected for the feature, but still dangerous if users are not clearly informed that their image/video data is being transmitted and processed externally.
-H 'Model: p-video-animate' \
-d '{
"input": {
"video": "https://api.pruna.ai/v1/files/source-video-abc123",
"image": "https://api.pruna.ai/v1/files/reference-image-def456",
"resolution": "720p",
"target_fps": "original",
The synchronous example submits externally hosted references for user media to the prediction API, again enabling remote processing of potentially sensitive assets. While aligned with the skill's purpose, it remains a true privacy/security concern absent mandatory disclosure and user approval.
-H 'Try-Sync: true' \
-d '{
"input": {
"video": "https://api.pruna.ai/v1/files/source-video-abc123",
"image": "https://api.pruna.ai/v1/files/reference-image-def456"
}
}'
This follow-on skill recommendation uses an unpinned package reference, enabling silent changes to downloaded content. In an agent ecosystem, recommended next steps can still trigger risky dependency retrieval and should be treated as supply-chain relevant.
The skill suggests installing an unpinned external package, which is a valid supply-chain weakness. If exploited, the agent may load altered instructions or code from the package maintainer account, registry, or distribution path.
This line includes another floating dependency install command. Such references undermine reproducibility and create an opening for malicious updates to be consumed automatically by users or agents following the skill instructions.
The recommended installation command does not lock the dependency version, making the behavior of the skill chain mutable and potentially attacker-influenced. This is especially risky because the installed skill may affect future prompts and API usage.
This final npx skills recommendation is also unpinned and therefore subject to upstream drift or compromise. A compromised dependency could alter agent guidance or execute unintended installation-time behavior.
No suspicious patterns detected.