knowledge-gate

Pass

Audited by Gen Agent Trust Hub on Apr 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill resolves and executes a local CLI tool (knowledge-gate) at runtime. It constructs shell commands to query a database and fetch entry details based on user input or file paths.
  • [DATA_EXFILTRATION]: The skill reads from a local database file (.knowledge/vault.db) and writes activity logs to a project-local JSONL file (tmp/vault-refs.jsonl) using shell redirection.
  • [PROMPT_INJECTION]: The skill includes instructions that mandate the agent treat external content (vault entries) as authoritative "MUST" or "MUST-NOT" rules. This creates an indirect prompt injection surface where the contents of the knowledge vault can influence the agent's behavior and decisions.
  • Ingestion points: Data enters the context from .knowledge/vault.db via the knowledge-gate CLI tool (SKILL.md).
  • Boundary markers: Absent; the agent is instructed to treat results as authoritative rules rather than potentially untrusted data.
  • Capability inventory: The agent can execute shell commands via the CLI tool and perform file system writes to the tmp/ directory (SKILL.md).
  • Sanitization: No sanitization or validation of the vault's content is described prior to interpolation into the agent's logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 9, 2026, 05:44 AM
Security Audit — agent-trust-hub — knowledge-gate