The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands such as ls, grep, and head across multiple local directories, including ~/Dev/, ~/bin/, and ~/.claude/. The instructions specify interpolating user-provided keywords into commands (e.g., grep -ril "[keywords]"), which could present a command injection risk if the agent does not properly sanitize the input.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and processing data from external files during its analysis phase.
Ingestion points: The skill reads the first 10-20 lines of existing markdown files in ~/.claude/skills/, ~/.claude/agents/, and ~/.claude/commands/ using head to assess functionality overlap.
Boundary markers: No delimiters or isolation instructions are provided to prevent the agent from following malicious instructions that might be embedded within the files being searched.
Capability inventory: The skill allows for the creation of new artifacts and further command execution based on the results of its analysis.
Sanitization: There is no evidence of sanitization or filtering of the content read from existing files before it is presented to the agent for decision-making.

creation-guard