intent-classifier
Pass
Audited by Gen Agent Trust Hub on Aug 23, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill functions as an intent classifier and router, which presents an attack surface for indirect prompt injection. Malicious user input could be designed to manipulate the classification algorithm or bypass routing logic to invoke unauthorized skills.
- Ingestion points: The skill processes natural language input to determine intent as described in SKILL.md.
- Boundary markers: The classification algorithm in references/classification-rules.md uses confidence thresholds and confirmation loops for low-confidence matches.
- Capability inventory: The skill can route to multiple high-privilege agents like security-code-auditor and skill-creator.
- Sanitization: The provided rules do not include specific input sanitization to prevent adversarial manipulation of the routing logic.
- [COMMAND_EXECUTION]: The skill provides a shell script at scripts/dynamic-catalog.sh for generating a skill registry. While primarily a development tool, it executes commands to scan the local file system and extract metadata from skill files.
Audit Metadata