The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided specifications and requirements to generate implementation plans. These plans contain shell commands and Python code intended to be executed by the agent or subsequent sub-skills (e.g., ltk:executing-plans). The lack of explicit boundary markers or input sanitization allows for a potential attack surface where malicious input could influence the generated commands or code.
Ingestion points: User-supplied task specifications and requirements provided during the planning phase (SKILL.md).
Boundary markers: No explicit delimiters or instructions to ignore embedded commands within the input data are defined.
Capability inventory: The skill instructions generate implementation plans involving file creation/modification, git commands, and pytest execution (SKILL.md).
Sanitization: No sanitization, validation, or escaping of the user-provided requirements is performed before interpolation into the plan template.
[COMMAND_EXECUTION]: The skill provides templates that generate shell commands for version control (git add, git commit) and for running tests (pytest). These are standard development operations and are explicitly documented for use within implementation plans.

writing-plans