The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute a command-line utility for project analysis. User-provided input from $ARGUMENTS is interpolated directly into the command string. Because this input is not properly escaped or sanitized, it allows for command injection if the arguments contain shell metacharacters like semicolons or pipes.
[DYNAMIC_EXECUTION]: The skill dynamically constructs shell commands and instructions for sub-agents at runtime by combining task data and user input. The lack of validation for these dynamic components introduces risks of arbitrary code execution, though it is central to the skill's logic for processing project-specific documentation needs.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves reading and summarizing local project files, which creates a surface for indirect prompt injection. Malicious instructions hidden in source code or existing project documentation could potentially influence the behavior of the documentation agents.\n- Ingestion points: Project source files read via Glob and Read tools during documentation tasks.\n- Boundary markers: None. The skill does not use specific delimiters to isolate untrusted content from agent instructions.\n- Capability inventory: The agents have access to Write, Bash, and spawn_agents_on_csv tools.\n- Sanitization: No filtering or validation is applied to findings extracted from project files before they are utilized by the LLM in subsequent documentation waves.

project-documentation-workflow