relationship-debrief
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data in the form of meeting transcripts located in
~/Documents/transcriptions/OpenOats/. These files contain verbatim text from external speakers which could include malicious instructions. - Ingestion points: Verbatim transcript sections in Markdown files.
- Boundary markers: The skill uses section headers (e.g.,
## Transcript) to identify data, but lacks explicit delimiters or instructions to ignore embedded commands within the transcript text. - Capability inventory: The skill performs file read and write operations within the
My Social Circle/andMy Meetings/directories. - Sanitization: There is no specific evidence of sanitization or filtering of speaker content before it is processed or used to draft CRM updates.
- [DATA_EXFILTRATION]: The skill accesses sensitive personal and professional data, including a private roster (
My Social Circle/Expo Team.md) and individual person files. While the skill's instructions focus on local file updates, the access to this data represents a potential exposure surface if the agent were manipulated into unauthorized output.
Audit Metadata