The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell-based tools (grep and git) to perform its primary functions. User-supplied input from task descriptions is interpolated into command arguments, which is a common pattern for developer tools.
Evidence: The duplication check in Step 2 uses grep with a user-derived keyword, and Step 7 executes git commit using the task title in the commit message.
[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it reads content from external files that may contain untrusted data.
Ingestion points: The skill reads jaan-to/roadmap.md, roadmap-add.learn.md, and system-level protocol documents.
Boundary markers: No explicit delimiters or specific instructions to ignore embedded commands are present in the skill definition to isolate data from instructions.
Capability inventory: The skill has capabilities to modify files via Edit and Write tools and interact with the git repository through shell commands.
Sanitization: No explicit content validation or escaping logic is defined for the data read from these external sources.

roadmap-add