computer-workflows

Pass

Audited by Gen Agent Trust Hub on Oct 31, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The 'File Organization' workflow ingests untrusted data by extracting text from external documents (e.g., PDFs) to identify candidates for relocation.
  • Ingestion points: references/file-organization/WORKFLOW.md (Step 3) utilizes pdftotext to process document content.
  • Boundary markers: None identified; instructions do not explicitly warn the agent to ignore instructions embedded within the processed text.
  • Capability inventory: The skill possesses capabilities to execute shell scripts (inventory.sh), move files, and delete system configurations (references/macos-complete-uninstall/WORKFLOW.md).
  • Sanitization: No sanitization or validation of the extracted text is performed before it is used to influence the agent's decision-making logic.
  • [DYNAMIC_EXECUTION]: The inventory.sh script (located in references/macos-complete-uninstall/scripts/inventory.sh) uses Bash heredocs to execute embedded Python source code at runtime for process ancestry analysis and browser profile inspection. This represents dynamic execution of code packaged within the skill.
  • [COMMAND_EXECUTION]: The skill uses various system utilities to perform an inventory and removal of software residues, including security dump-keychain for inspecting Keychain service names and sqlite3 for querying the macOS TCC (Transparency, Consent, and Control) database. These commands access sensitive local data stores to identify entries related to the target software.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 31, 2026, 03:09 AM