recall
Pass
Audited by Gen Agent Trust Hub on Nov 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests historical data from chat transcripts and external project records, creating a surface for indirect prompt injection where instructions embedded in past conversations or issue descriptions could influence current agent behavior.\n
- Ingestion points: Chat transcripts located at
~/.cursor/projects/<slug>/agent-transcripts/and shared records from source control, issue trackers, and chat channels.\n - Boundary markers: The skill relies on subagent filtering and a specific output contract, but lacks formal structural delimiters (like XML tags or multi-line markers) to isolate untrusted data from instructions.\n
- Capability inventory: The skill leverages
git,gh,ls, andgrepto retrieve and verify information.\n - Sanitization: Explicitly requires the agent to sanitize private context before generating public output.
Audit Metadata