The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install external command-line utilities from third-party sources using multiple package managers.
Evidence: cargo install toml-cli in SKILL.md and references/toml-cli.md.
Evidence: go install github.com/tmq/tmq@latest in SKILL.md and references/tmq.md.
Evidence: pip install yq and uvx yq in SKILL.md.
Evidence: npx @toml-tools/editor in SKILL.md.
[COMMAND_EXECUTION]: The skill includes workflows that modify sensitive system-level configuration files and interact with system services, operations which typically require elevated administrative privileges.
Evidence: Modifying /etc/pihole/pihole.toml in SKILL.md.
Evidence: Executing pihole restartdns in SKILL.md.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from external TOML files and uses that data in file-write and system-modification operations.
Ingestion points: Reads configuration data from files such as pyproject.toml and /etc/pihole/pihole.toml (SKILL.md).
Boundary markers: None present in the instructions to prevent the agent from following instructions embedded within the TOML data.
Capability inventory: The skill has the capability to write to files and execute shell commands (via allowed-tools: Bash).
Sanitization: No explicit sanitization or validation of the TOML content is performed before modification, beyond basic structure validation via tomllib.

toml-editor