The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill recommends a high-risk privilege escalation vector in its troubleshooting guide by suggesting the addition of 'NOPASSWD: ALL' to the sudoers configuration for the automation user (found in 'references/troubleshooting.md').
[DATA_EXFILTRATION]: The skill provides documentation for accessing highly sensitive system files, specifically referencing the usage and paths of SSH private keys such as '~/.ssh/id_rsa' (found in 'references/inventory.md').
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill instructs the agent to ingest and process external content from potentially untrusted files using modules like 'ansible.builtin.template', 'ansible.builtin.copy', and 'ansible.builtin.include_tasks' (found in 'references/playbooks.md' and 'references/modules.md').
Boundary markers: No instructions are provided to the agent to use delimiters or ignore embedded instructions within these external files.
Capability inventory: The skill documents extensive administrative capabilities across all referenced scripts, including arbitrary shell command execution ('shell', 'command'), script execution ('script'), container management ('docker_container'), and network operations ('uri').
Sanitization: The skill lacks any logic for sanitizing, escaping, or validating variables that are interpolated into templates or shell commands.
[COMMAND_EXECUTION]: The skill heavily documents the use of modules for arbitrary command execution on target systems, which is the primary function of the tool but requires careful oversight when executed by an AI agent.

ansible