pivot-check
Pass
Audited by Gen Agent Trust Hub on Jul 28, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill facilitates indirect prompt injection by instructing the agent to ingest and analyze all content within the memory directory (profiles, debriefs, learnings, channel maps) without implementing safety measures for untrusted data.\n
- Ingestion points: The instruction to Read EVERYTHING in memory/ in the 'Read first' section and evidence analysis in Phase 2.\n
- Boundary markers: None present. The skill does not provide delimiters or instructions to separate memory content from the agent's core instructions.\n
- Capability inventory: The skill includes instructions to write decision summaries to the filesystem (memory/decisions/pivot-check-.md).\n
- Sanitization: No sanitization, escaping, or filtering logic is specified for the ingested memory data.
Audit Metadata