skills/datadrivenconstruction/ddc_skills_for_ai_agents_in_construction/pandas-construction-analysis/Gen Agent Trust Hub
pandas-construction-analysis
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external construction data from various file formats. This creates a surface for indirect prompt injection where an attacker could embed instructions within a data file (e.g., in an Excel cell) to influence the agent's behavior.
- Ingestion points: The skill instructions and documentation describe reading data from user-provided files using
pandas.read_csv,pandas.read_excel, andpandas.read_json(instructions.md,SKILL.md). - Boundary markers: There are no explicit boundary markers or instructions in
instructions.mdthat direct the agent to treat all data from files as untrusted content or to ignore any linguistic instructions found within the data. - Capability inventory: The skill has filesystem permissions enabled to read input data and write output reports (
claw.json), providing the agent with the ability to modify local project files based on processed input. - Sanitization: No data sanitization or structural validation steps are defined in the instructions to prevent malicious content within the data from being interpreted as instructions.
Audit Metadata