vault-unlock
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions for the agent to execute a local Python script (
scripts/unlock.py) to process a binary file (vault.bin). - [CREDENTIALS_UNSAFE]: The script
scripts/unlock.pycontains a hardcoded passphrase (buddy-vault-2026) used for the PBKDF2 key derivation required to decrypt the vault file. This credential is functionally necessary for the skill and is managed by the vendor. - [INDIRECT_PROMPT_INJECTION]: The skill reads and decrypts data from an external binary file (
vault.bin) and instructs the agent to repeat the resulting output exactly to the user. This creates a potential attack surface where malicious instructions could be embedded in the encrypted data to influence the agent. - Ingestion points: The agent context ingests decrypted data from
vault.binprocessed byscripts/unlock.py. - Boundary markers: None. The instructions specifically state to "Podaj uzytkownikowi dokladnie to, co pojawilo sie na stdout" (Provide the user exactly what appeared on stdout).
- Capability inventory: The skill uses a Python script to read from the file system and output to the console.
- Sanitization: The decrypted secret is not sanitized, filtered, or validated before being presented to the agent or the user.
Audit Metadata