The Agent Skills Directory

[PERSISTENCE]: The skill appends user-defined shell functions to the ~/.zshrc file, ensuring that the created aliases persist across different terminal sessions and system reboots.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands like ls and grep to interact with the local file system and search through Claude session logs located in ~/.claude/projects/.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it incorporates untrusted user-provided content (search hints and alias names) directly into shell commands and file-writing operations without evident sanitization.
Ingestion points: User-provided strings for unique-content hints (used in a grep search) and short names (used as a shell function name) in SKILL.md.
Boundary markers: Absent; there are no instructions to delimit or sanitize these inputs to prevent shell command breakout.
Capability inventory: The skill uses ls, grep, and redirection (>>) to modify the ~/.zshrc configuration file in SKILL.md.
Sanitization: Absent; the skill does not specify any validation or escaping for the user-provided strings before they are used in sensitive shell operations.

find-session