decision-logger
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes meeting transcripts and decision logs which are external, untrusted data sources. Malicious instructions embedded in these files could potentially influence agent behavior when the history is loaded during Phase 1 of a board meeting.
- Ingestion points: The skill reads data from
memory/board-meetings/decisions.mdand variousYYYY-MM-DD-raw.mdfiles. - Capability inventory: The
scripts/decision_tracker.pyscript reads these files and outputs their contents to the agent's context for summarization and conflict detection. - Boundary markers: The skill relies on a structured markdown schema defined in
templates/decision-entry.md, but it does not include explicit prompt delimiters or instructions to the agent to ignore potentially adversarial content within the decision fields. - Sanitization: No sanitization or filtering of the decision content is performed before the data is presented to the agent.
Audit Metadata