capa-officer

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted CAPA record data from external JSON files or interactive input and includes it in reports that the agent interprets.
  • Ingestion points: The scripts/capa_tracker.py script reads data from external JSON files provided via the --capas command-line argument or through user input in --interactive mode.
  • Boundary markers: The script generates text reports and JSON summaries that include data from the input (such as titles, descriptions, and root cause statements) without using specific delimiters or instructions to prevent the agent from being influenced by embedded commands.
  • Capability inventory: The script performs file reading and standard output operations; while these are not inherently dangerous, the agent viewing the output may have higher-privileged capabilities that could be exploited by instructions hidden in the data.
  • Sanitization: The script does not perform sanitization, escaping, or validation on the ingested text fields before they are presented in the generated management reports.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 12:32 PM
Security Audit — agent-trust-hub — capa-officer