narrative-convergence

Pass

Audited by Gen Agent Trust Hub on Aug 25, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local executable to notify the operator of detected signals. While common for automation, this represents an external execution point.\n
  • Evidence: The bash command ./notify -f .pending-notify-temp/narrative-convergence-${today}.md is called in Step 7.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) because it aggregates data from multiple other skills that may process untrusted content from the web or social media.\n
  • Ingestion points: The skill reads files from output/.chains/*.md and memory/logs/*.md in Steps 1 and 2 to build its entity/theme map.\n
  • Boundary markers: No specific boundary markers or explicit instructions are provided to prevent the AI from obeying instructions embedded in the ingested content.\n
  • Capability inventory: The skill possesses the capability to write files (memory/topics/convergence-signals.md) and execute shell commands (./notify).\n
  • Sanitization: The skill lacks sanitization mechanisms for the data extracted from source files, allowing potentially malicious synthesized content to propagate to logs or notifications.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 25, 2026, 06:39 PM
Security Audit — agent-trust-hub — narrative-convergence