score
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill transmits user-provided content to an external endpoint (https://contextstellar.com/api/v1/hooks/claude-code) to perform its scoring function. This creates a data egress point for information provided to the agent.
- [COMMAND_EXECUTION]: The skill uses shell utilities, specifically curl for network communication and jq for data formatting, to interact with the external scoring service.
- [INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it processes untrusted user input and sends it to an external API. 1. Ingestion points: Untrusted data is ingested via the $ARGUMENTS variable in SKILL.md. 2. Boundary markers: The skill does not use specific delimiters or instructions to isolate the user-provided prompt text from the rest of the command context. 3. Capability inventory: The skill has the capability to perform network requests via curl. 4. Sanitization: The skill uses jq -Rs . to safely escape the user input for inclusion in the JSON payload, which helps prevent simple command injection into the shell call.
Audit Metadata