Video Editing Agent (VEA)
ReviewAudited by ClawScan on May 10, 2026.
Overview
VEA appears to be a legitimate local video-editing tool, but it asks for broad Google and third-party credentials and includes shell helpers that need careful handling.
Install only if you are comfortable giving this local VEA service access to the listed media/AI provider credentials. Use least-privilege keys or a dedicated Google project, confirm each video edit before processing, and clean up local indexing/output folders after sensitive jobs.
Findings (6)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
The local VEA server may be able to act using your Google and media-service credentials, not just the declared Memories.ai key.
The skill asks for multiple provider credentials and Google application-default credentials, but the artifact does not explain Google OAuth scopes, account boundaries, or revocation/containment.
gcloud auth application-default login # Authenticate GCP ... API Keys (in `config.json`): `MEMORIES_API_KEY`, `GOOGLE_API_KEY`, `ELEVENLABS_API_KEY`, `SOUNDSTRIPE_KEY`
Use least-privilege API keys, prefer dedicated throwaway service accounts/projects, avoid broad personal Google ADC where possible, and revoke keys when finished.
A malicious or malformed server path/port value could potentially cause extra shell commands to run locally.
VEA_DIR and PORT are interpolated into a shell command string executed by tmux without validation or escaping; if those values come from untrusted input, shell metacharacters could run unintended commands.
"cd $VEA_DIR && source .venv/bin/activate && python src/app.py --port $PORT; read"
Only use trusted numeric port values, do not pass user-supplied shell text into this script, and harden the script by validating PORT and safely quoting/escaping VEA_DIR.
You are trusting the external repository and installer at install time.
The setup pulls code and an installer from external sources. This is expected for an open-source local service, but the commands are not pinned in the provided artifact.
git clone https://github.com/Memories-ai-labs/vea-open-source.git ... curl -LsSf https://astral.sh/uv/install.sh | sh ... uv sync
Review the repository, pin commits/versions when possible, and avoid running remote install scripts without inspection.
Selected video content and generated text may leave your machine for third-party AI services.
The artifact discloses provider data flows involving video-derived content and narration text, which are purpose-aligned but privacy-relevant.
Video frames sent to Memories.ai for AI comprehension; ElevenLabs receives text for TTS narration
Do not process sensitive videos unless you accept those providers' data handling terms; use project-specific keys and review provider retention settings.
Video-derived metadata and generated content may remain on disk after the editing task.
The skill stores video indexes, clip plans, narration, subtitles, and outputs locally for reuse in later editing steps.
Creates `~/vea/data/indexing/PROJECT_NAME/media_indexing.json` ... `clip_plan.json` ... `narrations/` ... `subtitles/`
Review and delete `~/vea/data/indexing/` and `~/vea/data/outputs/` when projects are complete, especially for private videos.
The local server may keep running after the immediate request finishes.
The helper can start the VEA server as a background process. This is purpose-aligned for long video jobs, and the same helper documents a stop command.
nohup ./.venv/bin/python -m src.app > /tmp/vea.log 2>&1 &
Stop the server when done and review `/tmp/vea.log` if troubleshooting or handling sensitive projects.