use-claude
Pass
Audited by Gen Agent Trust Hub on Jul 17, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill describes how to invoke the
claudeCLI tool using shell commands. It also provides a Python one-liner andjqexamples to parse JSON output from a temporary file (/tmp/claude-out.json). - [INDIRECT_PROMPT_INJECTION]: The skill facilitates the delegation of context to an external agent, which creates a potential surface for indirect prompt injection if the source data is untrusted.
- Ingestion points: Untrusted data enters the workflow via the prompt argument in the
claude -p "<prompt>"command. - Boundary markers: The instructions do not mandate specific delimiters or boundary markers to separate system instructions from user-provided context in the delegated prompt.
- Capability inventory: Depending on invocation flags, the delegated process can perform file reads, directory searches, and file edits (
--permission-mode acceptEdits). - Sanitization: The provided examples do not include explicit sanitization or escaping of the content interpolated into the prompt string.
Audit Metadata