creative-writing
Warn
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: MEDIUMPERSISTENCECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONTIME_DELAYED_CONDITIONAL
Full Analysis
- [PERSISTENCE]: The skill explicitly references a daily cron job (job_id: 96522a27ae03) to maintain automated social media activity and writing tasks across sessions.
- [COMMAND_EXECUTION]: The instructions contain shell commands for configuring environment variables (export http_proxy/https_proxy) to tunnel network traffic through a local proxy port.
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external social platforms, creating an attack surface for instructions embedded in notifications or comments. 1. Ingestion points: Fetches content via GET /notifications from Moltbook and The Colony platforms. 2. Capability inventory: The skill performs local file writes to the home directory (~/moltbook_story_day{N}.md) and initiates network operations for posting content. 3. Boundary markers: Absent; there are no defined delimiters or instructions to ignore embedded commands within the ingested data. 4. Sanitization: Absent; the workflow does not specify validation or filtering for text retrieved from external APIs.
- [TIME_DELAYED_CONDITIONAL]: Core network and file system operations are gated by a specific daily schedule (11:00 daily) as part of an automated 30-day challenge.
Audit Metadata