writing-cli-skills
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONPRIVILEGE_ESCALATION
Full Analysis
- [PRIVILEGE_ESCALATION]: The template provided in 'references/template.md' includes a command example using 'sudo' for Linux package installation ('sudo apt install tool-name'). While common for software installation, it provides a pattern for executing commands with elevated privileges.
- [INDIRECT_PROMPT_INJECTION]: The skill's 'Quick Start' section in 'SKILL.md' suggests using 'WebFetch' on official documentation and man pages to gather information. This introduces a vector for processing untrusted external content.
- Ingestion points: External documentation URLs via 'WebFetch' described in 'SKILL.md'.
- Boundary markers: No guidance is provided on using delimiters or safety instructions to prevent the agent from obeying instructions embedded in the fetched documentation.
- Capability inventory: The skill involves administrative tasks like creating symbolic links and managing directory structures for skill deployment.
- Sanitization: The guide does not include steps for sanitizing or validating external content before it is incorporated into the agent's workspace.
Audit Metadata