pp-mercury
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install CLI binaries from the '@mvanhorn/printing-press-library' NPM package and the 'mvanhorn/printing-press-library' GitHub repository.
- [COMMAND_EXECUTION]: The skill functions by executing shell commands via the 'mercury-pp-cli' binary to manage bank accounts and process transactions.
- [DATA_EXFILTRATION]: The '--deliver webhook:' feature allows the output of any command (which may include sensitive banking and organization data) to be sent to a user-specified external URL.
- [INDIRECT_PROMPT_INJECTION]: The skill reads and processes external data from the Mercury API (transactions, invoices, and customers) and a local SQLite database, creating a surface for potential indirect prompt injection attacks.
- Ingestion points: Data retrieved from Mercury API accounts, transactions, and bookkeeping endpoints.
- Boundary markers: No explicit prompt boundary markers or instruction-isolation warnings are provided.
- Capability inventory: The skill has capabilities for file writing ('workflow archive'), network communication (API access and webhooks), and command execution.
- Sanitization: No evidence of sanitization or filtering for incoming API data is documented in the skill instructions.
Audit Metadata