using-program-management
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script for delivery forecasting purposes.\n
- Evidence: The file
estimation-and-forecasting.mdprovides the specific shell commandpython plugins/axiom-program-management/scripts/forecast.py observed-throughput.csv --remaining 20 --seed 0to perform Monte Carlo simulations.\n- [INDIRECT_PROMPT_INJECTION]: The skill's workflow involves ingesting data from external sources, which creates a potential surface for indirect prompt injection attacks.\n - Ingestion points: As described in
SKILL.mdandestimation-and-forecasting.md, the skill reads project tracker records, milestones, release evidence, and throughput data from CSV files.\n - Boundary markers: No specific delimiters or instructions to treat external data as untrusted or ignore embedded instructions are present in the guidelines.\n
- Capability inventory: The skill allows for local command execution (
python) and interaction with project trackers, which could be misused if injected data influences the agent's logic.\n - Sanitization: The documentation does not outline procedures for sanitizing or validating the content of the ingested project data or CSV files.
Audit Metadata