debug
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides multiple bash commands utilizing the
rg(ripgrep) tool to search log files based on user-supplied identifiers like issue IDs and session keys. - [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read application logs, which are a source of untrusted data originating from the application's runtime and external interactions.
- Ingestion points: The skill reads from
log/symphony.logand rotated log variants (log/symphony.log*) as specified in SKILL.md. - Boundary markers: No delimiters or explicit instructions to ignore potentially malicious embedded content within the logs are provided.
- Capability inventory: The skill leverages the agent's capability to execute shell commands (specifically
rg) on the ingested data. - Sanitization: There is no specified sanitization or validation of the log content or the identifiers before they are used in commands.
Audit Metadata