spec-to-code-compliance
Warn
Audited by Gen Agent Trust Hub on May 16, 2026
Risk Level: MEDIUMPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [PROMPT_INJECTION]: The skill metadata includes 'verified: true' and 'lastVerifiedAt' fields which are deceptive. These mimic official security verification markers and may cause misjudgment of the skill's safety, especially since the code is currently only a scaffold.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. It is designed to read and analyze implementation code and specifications using 'Read' and 'Grep' tools. There are no boundary markers or instructions to ignore instructions within the analyzed data. The agent has access to powerful tools like 'Bash' and 'WebFetch' that could be subverted if the input data contains malicious payloads.
- [DATA_EXFILTRATION]: The skill possesses access to the 'WebFetch' and 'Bash' tools. Although no malicious network operations are hardcoded in the scaffold, the instructions to 'Perform the skill's main function using available tools' provide a vector for data exfiltration of any context the agent gathers from the local filesystem.
Audit Metadata