The Agent Skills Directory

[PROMPT_INJECTION]: The skill identifies a potential indirect prompt injection surface and includes proactive guardrails to mitigate it.
Ingestion points: The agent retrieves issues, comments, and documents from the Linear platform, which are sources of untrusted external data.
Boundary markers: METHODOLOGY.md includes a dedicated 'Trust boundary' section that explicitly instructs the agent to treat project content as data, not instructions, and to never follow embedded requests or run commands found within them.
Capability inventory: The skill uses Linear MCP tools for project management and the curl command for uploading file attachments.
Sanitization: The instructions emphasize verifying that changes follow confirmed project scope and surfacing conflicts to the user rather than executing them.
[COMMAND_EXECUTION]: The methodology instructs the use of the curl command specifically for the Linear attachment upload flow.
Evidence: Section 10 of METHODOLOGY.md details a multi-step process for performing an HTTP PUT of raw bytes to a signed URL provided by the Linear API.
Context: This is a documented and standard procedure for handling file attachments in the Linear ecosystem and is confined to project-tracking evidence like screenshots.

linear-monitor