The Agent Skills Directory

[PROMPT_INJECTION]: The skill is subject to indirect prompt injection because it ingests and processes content from untrusted local markdown files. * Ingestion points: The tool reads all files matching the 'ADR-*.md' pattern within a user-defined directory. * Boundary markers: There are no delimited boundaries or specific instructions to the agent to disregard potential commands embedded within the ADR text. * Capability inventory: The skill's permissions are limited to reading local files and writing a markdown roadmap file; it cannot access the network or execute system commands. * Sanitization: The skill does not perform any validation or sanitization of the input text before including it in the logical processing for roadmap generation.

adr-roadmap