The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses ls to scan directory structures and curl to interact with a Squad project's API. These operations are scoped to project-level knowledge synthesis and are benign in this context.
[EXTERNAL_DOWNLOADS]: Uses curl to fetch data from $BASE_URL/api/. These are internal API endpoints defined in the project's shared configuration and are used for reading project metadata and task details.
[DATA_EXFILTRATION]: While the skill reads sensitive project data (architecture, decisions, board state), the data is only used to generate local markdown files in the wiki/ directory. No evidence of sending project data to unauthorized external domains was found.
[CREDENTIALS_UNSAFE]: The skill references an AUTH_HEADER for API requests. According to the instructions, this header is expected to be provided by the environment/shared configuration (../squad/shared.md), following safe credential management practices.
[PROMPT_INJECTION]: The skill includes instructions to ignore certain card fields (like impl_notes) to prevent noise, but no malicious pattern attempting to override the agent's core safety guidelines was detected.

squad-gen-wiki