infrahub-analyzing-diagnostics
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from log files and manifests which may contain malicious instructions intended to manipulate agent responses.
- Ingestion points: Uses Read, Grep, and Glob tools to access user-specified bundle directories.
- Boundary markers: The skill requires excerpts to be quoted and linked to specific file paths.
- Capability inventory: Includes the ability to perform WebFetch and execute CLI commands via Bash.
- Sanitization: Explicit instructions command the agent to mask credentials and strip volatile tokens like IPs and UUIDs before processing.
- [COMMAND_EXECUTION]: The skill executes gh CLI commands via Bash using search keys derived from log tracebacks. This introduces a surface where malformed log content could influence command arguments, though the skill provides normalization rules to mitigate this.
Audit Metadata