openclaw-executive-assistant-local

Pass

Audited by Gen Agent Trust Hub on Sep 12, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONPERSISTENCEPROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external sources, creating a surface for indirect prompt injection attacks.
  • Ingestion points: Files placed in code-along/01-data-intake-review/incoming/ and .eml files in code-along/03-offline-communications-triage/eml/ are read by the AI agent.
  • Boundary markers: The provided prompts do not include explicit delimiters or instructions to ignore potentially malicious embedded content within the source files.
  • Capability inventory: The skill uses a hypothetical AI CLI tool to read these files and write markdown reports to local storage (outputs/).
  • Sanitization: There is no evidence of content sanitization or validation before the data is processed by the LLM.
  • [PERSISTENCE]: The documentation provides instructions for maintaining access and automated execution across sessions.
  • Evidence: The skill includes examples of scheduling daily and weekly tasks using the system cron facility (e.g., 0 18 * * * cd ~/openclaw-assistant && ./generate-daily-log.sh).
  • [PROMPT_INJECTION]: The skill uses override-style keywords to dictate agent behavior, which matches common prompt injection patterns.
  • Evidence: The Troubleshooting section suggests using the prefix CRITICAL: Output must be valid Markdown... to override potential formatting inconsistencies in the agent's output.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 12, 2026, 06:50 PM
Security Audit — agent-trust-hub — openclaw-executive-assistant-local