copilot-money

Fail

Audited by Gen Agent Trust Hub on Oct 26, 2026

Risk Level: HIGHCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill performs automated extraction of authentication tokens (refresh tokens) from browser IndexedDB storage, including Chrome, Safari, Firefox, and Arc. These secrets are stored locally at ~/.config/copilot-money/config.json for subsequent access.
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of the copilot-money-cli package from the Python Package Index. This is explicitly labeled as an unofficial third-party utility that is not affiliated with the financial service provider.
  • [COMMAND_EXECUTION]: The skill relies on executing various shell commands to query financial data, synchronize bank connections, and manage local configurations.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external transaction, holding, and account data which could potentially be manipulated by an attacker to include malicious instructions intended for the agent.
  • Ingestion points: Financial data retrieved from the API via copilot-money transactions and copilot-money holdings (file: SKILL.md).
  • Boundary markers: None identified; the skill does not appear to use delimiters or warnings to isolate external data from instructions.
  • Capability inventory: Shell command execution for querying balances, transactions, and net worth.
  • Sanitization: There is no evidence of input sanitization or validation performed on the retrieved financial data before it is processed by the agent.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Oct 26, 2026, 09:37 AM