The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill includes a bootstrap script (scripts/vss-bootstrap.sh) that ensures the application source is available for execution.
The script attempts to detect a local installation or a previous clone in a cache directory.
If not found, it performs a shallow, sparse, and single-branch clone of https://github.com/open-edge-platform/edge-ai-libraries.git from the main branch. This is a vendor-owned resource matching the skill's author.
[COMMAND_EXECUTION]: The skill requires the execution of several shell commands to manage the environment and interact with the VSS API.
Executes bash to run the bundled bootstrap script.
Utilizes git for environment detection and source code acquisition.
Uses curl to perform file uploads and API queries against the Pipeline Manager.
Employs jq for parsing and transforming JSON responses from the backend.
[INDIRECT_PROMPT_INJECTION]: The skill processes structured data from external API responses, which represents a surface for indirect prompt injection if the backend data is compromised.
Ingestion points: The skill reads video metadata (filenames, tags) and search results (locators, scores) from the /manager/videos and /manager/search/query endpoints as described in SKILL.md and references/search-request.md.
Boundary markers: No explicit instructions are provided to the agent to ignore or delimit instructions found within these metadata fields.
Capability inventory: The skill has access to bash, curl, git, and jq for file system and network operations as detailed in SKILL.md and scripts/vss-bootstrap.sh.
Sanitization: The skill uses jq to extract specific fields, which provides some structural validation, but it does not explicitly sanitize the content of strings (like filenames) before they are interpolated into the prompt or subsequent shell commands.

vss-search-index