ingest-substack-csv

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from CSV files stored in inbox/substack-stats/. While it performs structural schema validation, the content of fields such as post titles or slugs is not sanitized, which could potentially influence agent behavior if they contain prompt-like instructions.
  • Ingestion points: CSV files located in the inbox/substack-stats/ directory, specifically reading the header and data rows for posts and subscriber aggregates (SKILL.md).
  • Boundary markers: Absent. The instructions do not specify the use of delimiters or specific commands to treat the CSV content as passive data only.
  • Capability inventory: The skill utilizes file system operations including reading files, writing warning stubs to ops/growth-analyst/, and moving successful exports to corpus/stats/. No network or shell execution capabilities are present.
  • Sanitization: The skill performs header validation against .schema-expected.json to detect format changes, but it does not describe sanitization or filtering of the actual text content within the data rows.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 02:36 PM
Security Audit — agent-trust-hub — ingest-substack-csv