fixing-broken-links
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The instructions direct the agent to execute shell commands like
curland[ -f ]using strings extracted from potentially untrusted files. This pattern is vulnerable to command injection if a URL or path contains shell metacharacters that are not properly escaped by the agent's execution environment. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from local files and external websites.
- Ingestion points: All URLs and local file paths extracted from target files as described in the extraction workflow in
SKILL.md. - Boundary markers: No explicit delimiters or instructions are provided to distinguish between data and potential instructions within the links or the search results used to fix them.
- Capability inventory: The skill possesses network capabilities (
curl), file system inspection ([ -f ]), and file writing permissions to update local content. - Sanitization: The instructions do not include steps to validate or sanitize extracted URLs before they are passed to shell commands or used in search queries.
Audit Metadata