The Agent Skills Directory

[DATA_EXFILTRATION]: The skill exposes a hardcoded absolute file path in the <execution_context> block: C:/Users/nicol/.codex/get-shit-done/workflows/map-codebase.md. This reveals internal environment information, specifically the local user's name ('nicol') and the specific directory structure of the developer's machine.
[COMMAND_EXECUTION]: The skill orchestrates the creation of a directory structure (.planning/codebase/), spawns multiple sub-agents (gsd-codebase-mapper) using the spawn_agent tool, and executes a commit action ('Commit codebase map'). These operations interact directly with the local filesystem and process management.
[PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by ingesting and processing untrusted data from the codebase being analyzed.
Ingestion points: Untrusted codebase files (brownfield codebases) and the {{GSD_ARGS}} user input parameter in SKILL.md.
Boundary markers: Absent. The instructions do not provide delimiters or 'ignore' directives to prevent sub-agents from obeying instructions embedded within the codebase files they are mapping.
Capability inventory: Spawning agents, writing files to the disk, and performing git commits.
Sanitization: Absent. The {{GSD_ARGS}} variable is interpolated directly as a 'Focus area' without validation or escaping logic.

gsd-map-codebase