dessix-skill
Pass
Audited by Gen Agent Trust Hub on Feb 26, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes content from local Dessix workspace blocks, creating an attack surface for indirect prompt injection.\n
- Ingestion points: Untrusted content enters the agent context through the
dessix_read_block,dessix_get_skill, anddessix_get_topic_contexttools executed viascripts/dessix-bridge.mjs.\n - Boundary markers: Absent. The skill instructions in
SKILL.mddo not provide delimiters or warnings to treat block content as untrusted data.\n - Capability inventory:
scripts/dessix-bridge.mjsfacilitates reading, searching, and writing (creating, updating, deleting) blocks in the local Dessix application vianet.createConnection.\n - Sanitization: Absent. No filtering or validation of the retrieved block content is performed by the bridge script or skill instructions.\n- [SAFE]: No data exfiltration detected. The bridge client communicates strictly with local endpoints, such as
~/.dessix/mcp/dessix-mcp-bridge.sockor\\\\.\\pipe\\dessix-mcp-bridge.\n- [SAFE]: No remote code execution or unverifiable dependencies were found. The skill uses a local script that relies exclusively on built-in Node.js modules.
Audit Metadata