The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection when processing Linear issue descriptions. The 'VantageEx' epic format involves parsing structured markdown sections (Objective, Skills, Repos) that are user-authored. Malicious instructions placed in these sections could influence the agent's behavior when auditing or grooming epics.
Ingestion points: Linear issue titles and descriptions (read in scripts/0.1.0/linear.nu and references/audit-checklist.md).
Boundary markers: None present to distinguish data from instructions.
Capability inventory: Network access via http post to Linear API, file system access (implied for script execution), and issue/comment management.
Sanitization: Absent for parsed markdown sections.
[COMMAND_EXECUTION]: The Nushell client scripts/0.1.0/linear.nu constructs GraphQL queries and mutations by interpolating variables directly into string templates. This pattern allows for GraphQL injection if inputs like $title or $filter contain characters that break the GraphQL syntax. For instance, in linear create-issue, the $title variable is inserted into the mutation string without escaping.
[EXTERNAL_DOWNLOADS]: The skill documentation recommends setting up the Linear MCP server from https://mcp.linear.app/mcp. Linear is a well-known project management service, and this connection is a standard requirement for the skill's intended functionality.

linear