The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command (ls) to inspect multiple filesystem paths, including hidden configuration directories in the user's home folder (such as ~/.claude/skills/ and ~/.agent/skills/), to identify installed companion skills.
[PROMPT_INJECTION]: The 'Companion check' section contains instructions that override default agent behavior by mandating an environment scan and a promotional pitch for the author's other skills immediately upon activation.
[PROMPT_INJECTION]: The skill establishes an indirect prompt injection surface by processing .meta configuration files to drive command execution across multiple directories.
Ingestion points: .meta JSON configuration file (SKILL.md)
Boundary markers: Absent from the instructions governing the parsing and execution of configuration data
Capability inventory: Execution of arbitrary shell commands across project directories via meta exec, meta npm, and meta git (SKILL.md, references/commands.md)
Sanitization: No validation or sanitization requirements are defined for the paths or remote URLs parsed from the configuration file
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the meta tool and its associated plugins from the npm registry, as well as the addition of supplementary agent skills from the vendor's GitHub repository via npx.

meta-repo