land-assembly-expert

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • REMOTE_CODE_EXECUTION (MEDIUM): The skill modules (budgeting.py, phasing.py) and the main script (land_assembly_calculator.py) use sys.path.insert(0, ...) to dynamically include a directory located four levels above the skill's installation path (../../../../Shared_Utils). This pattern of dynamic loading from computed paths is risky; an attacker who can place files in that relative path on the host system could achieve local code execution when the calculator is run.
  • PROMPT_INJECTION (LOW): The skill is susceptible to indirect prompt injection (Category 8). It ingests untrusted data from a JSON file and interpolates fields like project_name and parcel address directly into Markdown reports. If these reports are subsequently read by another AI agent, malicious instructions embedded in the JSON could influence that agent's subsequent actions.
  • Ingestion points: load_and_validate_input in validators.py reads user-controlled JSON.
  • Boundary markers: Absent. Data is directly interpolated into Markdown strings in output_formatters.py and other modules.
  • Capability inventory: The script has file-write capabilities (save_markdown_report, save_json_output) and executes local imports based on path manipulation.
  • Sanitization: No sanitization or escaping is performed on the input data before it is written to the output reports.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 06:29 PM