scheduled-summary
Pass
Audited by Gen Agent Trust Hub on Aug 15, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from session databases, cron outputs, and log files to generate summaries. While it performs truncation and basic formatting, it lacks robust sanitization or security boundaries to prevent malicious instructions embedded in these sources from influencing an agent that subsequently reads the generated summary.
- Ingestion points: Data is ingested via SQLite databases (
--sessions-db), log files (--log-file), and directories containing JSON or Markdown files (--cron-dir,--memory-dir,--decisions-dir) inscripts/summarize.py. - Boundary markers: The output uses standard Markdown headings and bullets to separate sections. No explicit security delimiters or 'ignore embedded instructions' warnings are applied to the extracted content.
- Capability inventory: In
scripts/summarize.py, the skill is limited to local file system read access, SQLite query execution, and writing the summary to a local file. It does not possess network access, shell execution capabilities, or dynamic code evaluation features. - Sanitization: Content is normalized for whitespace and truncated to specific length limits (e.g., 60-180 characters) to reduce verbosity, which offers limited protection against semantic injection attacks.
Audit Metadata