skills/mukul975/anthropic-cybersecurity-skills/building-vulnerability-exception-tracking-system/Gen Agent Trust Hub
building-vulnerability-exception-tracking-system
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides a mechanism to ingest and store user-supplied text (justifications, compensating controls) in a database. This data is intended to be reviewed by humans or potentially automated systems, creating a surface for indirect prompt injection if the content is processed by an LLM without proper sanitization.
- Ingestion points: The
scripts/process.pyscript ingests exception request data from JSON files, and the Flask API examples inSKILL.mdprocess JSON payloads. - Boundary markers: No explicit delimiter or instruction-filtering logic is present in the data ingestion or storage functions.
- Capability inventory: The skill performs SQLite database writes and can send notification payloads to external Slack webhooks via the
requestslibrary. - Sanitization: The implementation correctly uses parametrized SQL queries to prevent SQL injection, but does not provide specific sanitization or validation for natural language instructions within the text justification fields.
Audit Metadata