The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands like ls, cat, mkdir, and node to handle file operations and metadata management.
[DATA_EXFILTRATION]: In resume mode, the skill executes cat on a path constructed using user-provided input (${THREAD_NAME}). This presents a path traversal risk if the agent does not sanitize the input to prevent access to files outside the intended .planning/threads/ directory.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it loads the content of existing thread files into the agent's context without using boundary markers.
Ingestion points: Contents of files in .planning/threads/ read via the cat command in the resume mode.
Boundary markers: None identified in the process instructions.
Capability inventory: Shell command execution (ls, cat, mkdir), file writing (cat >), and execution of a local Node.js script (gsd-tools.cjs).
Sanitization: No sanitization or validation of the file content is performed before it is ingested into the active session context.

gsd-thread