The Agent Skills Directory

[COMMAND_EXECUTION]: The skill generates shell commands such as git checkout -b {BRANCH_NAME} using values provided in the --branch argument. Because explicit branch names are not sanitized before being placed into the command template, this creates a command injection vulnerability if the agent executes the planned tasks.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from a user-specified file to perform task analysis and breakdown. This processed content could contain malicious instructions designed to manipulate the agent's behavior. 1. Ingestion points: The file path provided as a required argument is read in Phase 1 and Phase 5. 2. Boundary markers: The skill does not use delimiters or warnings to isolate external data from its own instructions. 3. Capability inventory: The skill utilizes Write, Edit, Bash (for git/gh), and Task (subagents) tools. 4. Sanitization: No sanitization is performed on the file content before it is processed by the agent or subagents.
[DYNAMIC_EXECUTION]: The skill dynamically constructs implementation plans and shell commands at runtime by concatenating user input with subagent results. These instructions are persisted to the filesystem to be followed by the agent in subsequent steps.

todo-task-planning