The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the data-liberation CLI tool to perform its primary tasks. It passes user-supplied URLs and local directory paths as arguments to this command, which could be exploited if the underlying tool does not adequately sanitize inputs against shell injection or SSRF.
[DATA_EXFILTRATION]: The publish command allows the agent to upload the contents of the local run directory to an external service (spacefast). Although this is the stated purpose of the skill, it represents a mechanism for transferring local data to a remote host.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from arbitrary external URLs, creating a surface for indirect instructions to influence agent behavior. 1. Ingestion points: Untrusted data enters the agent context through the data-liberation inspect and liberation commands referenced in SKILL.md. 2. Boundary markers: The instructions lack explicit delimiters or 'ignore' directives to help the agent differentiate between its core instructions and the content of the captured websites. 3. Capability inventory: The skill possesses capabilities for local file system writes, CLI execution, and network exfiltration via the publishing command. 4. Sanitization: There are no procedures defined in the skill instructions to filter or sanitize the content retrieved from the web before it is processed.

liberate