The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell utilities including find, sed, grep, cut, and xargs to discover files and extract metadata from the local filesystem. Specifically, it executes these commands over ~/.claude subdirectories to build a registry of available tools.
[COMMAND_EXECUTION]: The skill incorporates user-supplied input through the $ARGUMENTS variable to define the <plan_path>. If this variable is not properly sanitized before being used in filesystem operations or shell commands, it could lead to command injection.
[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection (Category 8) because it ingests and processes untrusted data from project files.
Ingestion points: The skill reads spec.md, prd.json, and brainstorm.md from the provided plan folder.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore or isolate embedded instructions within the processed files.
Capability inventory: The skill has the capability to execute shell commands, write updates to the local filesystem (updating spec.md and prd.json), and invoke other agents based on the content of the ingested files.
Sanitization: No sanitization or validation of the input text from the plan files is performed before it is used to match skills or influence agent behavior.

deepen-plan