The Agent Skills Directory

[DATA_EXFILTRATION]: The skill's primary function is to transmit project data to the external service handover.sh. It instructs the agent to attach original files such as SQL, HTML, JSON, and code, and to provide evidence including paths, URLs, and query outputs. While the skill requires authentication and mandates user approval for public visibility, the workflow involves moving potentially sensitive local context to a third-party platform.
[COMMAND_EXECUTION]: The skill utilizes shell commands to interact with the Handover service when the MCP interface is unavailable. Specifically, it executes handover create and handover publish to upload artifacts and directory contents.
[EXTERNAL_DOWNLOADS]: The instructions direct the agent to point users toward https://handover.sh/install for downloading the CLI tool if it is not already configured in the environment.
[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because it processes untrusted data from the active session to build continuation records.
Ingestion points: The skill gathers the objective, current state, decisions, and evidence (paths, URLs, queries, outputs) from the task context.
Boundary markers: There are no explicit delimiters or instructions to ignore embedded commands within the ingested data defined in the skill.
Capability inventory: The skill uses handover.create and handover.publish (via CLI or MCP) to transmit the constructed record to the external service.
Sanitization: No explicit sanitization or filtering of the content is performed before it is included in the published record.

handover-publish