feedback
Pass
Audited by Gen Agent Trust Hub on Aug 18, 2026
Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFLLTRATION]: The skill instructions require the agent to quote rules verbatim into a shared realm (@nks/feedback) to provide context for feedback. While this is the intended primary purpose of the skill, it could lead to the exposure of internal skill logic or system prompts to any user or agent with access to the shared feedback realm.
- [PROMPT_INJECTION]: The skill uses
nks_semantic_searchto ingest data from the shared@nks/feedbackrealm to check for duplicate entries. This represents an indirect prompt injection surface where maliciously crafted feedback entries from other users could influence the agent's logic during the comparison and categorization phase. - Ingestion points:
nks_semantic_searchqueries against the@nks/feedbackrealm in SKILL.md. - Boundary markers: None. The agent is instructed to read the field and 'add novelty to the node that stands'.
- Capability inventory: Writing to the
@nks/feedbackrealm using platform tools. - Sanitization: No explicit sanitization or filtering of the retrieved semantic search results is performed before the agent processes them.
Audit Metadata