The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses powerful shell commands to manage files, specifically rm -rf for uninstallation and unzip for installation. While these are restricted to the ~/.claude/skills/ and project directories, they represent a significant capability for modifying the agent's environment.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) because it extracts and displays text from the description and SKILL.md files of other installed skills.
Ingestion points: The scripts/list_skills.sh script and SKILL.md workflow read content from other skills' SKILL.md files using sed and cat.
Boundary markers: No specific delimiters or instructions are used to tell the agent to ignore embedded instructions within the processed skill descriptions.
Capability inventory: The agent has the ability to execute file operations (rm, cp, unzip, ls) and read file contents across the local filesystem.
Sanitization: The script uses basic string manipulation (sed, cut, tr) to extract metadata but does not perform any sanitization to prevent the interpretation of malicious instructions contained within those fields.

skill-manage