skills/dannote/dot-pi/session-reflect/Gen Agent Trust Hub

session-reflect

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local TypeScript helper script (scripts/session-db.ts) via tsx to manage data processing. This script ingests session logs into a local DuckDB instance and runs SQL queries to extract behavioral evidence.
  • [PROMPT_INJECTION]: The skill processes session logs which are a source of untrusted data, creating an indirect prompt injection surface. The instruction set mitigates this by explicitly directing the agent to treat log content as evidence for analysis rather than as new instructions. The helper script also implements a basic SQL escaping utility (sqlString) to sanitize text filters derived from user input.
  • [SAFE]: The skill operates entirely within the local environment, specifically targeting the user's session log directory (~/.pi/agent/sessions) for read access and a dedicated cache directory (~/.pi/agent/cache/session-reflect/) for its internal database. No network exfiltration or credential harvesting patterns were observed.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 10:36 PM
Security Audit — agent-trust-hub — session-reflect