skills/mr-kelly/skills/kelly-lesson/Gen Agent Trust Hub

kelly-lesson

Pass

Audited by Gen Agent Trust Hub on Aug 25, 2026

Risk Level: SAFEPROMPT_INJECTIONCREDENTIALS_UNSAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from curriculum materials and teacher-drafted lesson plans. \n
  • Ingestion points: The scripts/ingest_plan.mjs script parses external JSON and document data into structured records. \n
  • Boundary markers: Instructions direct the agent to analyze content rather than execute it, and data is partitioned into specific JSON schema fields. \n
  • Capability inventory: The skill can write to the Busabase API and perform local file exports via scripts/export_plans.mjs. \n
  • Sanitization: UI rendering logic in lesson-views.js utilizes escapeHtml to prevent cross-site scripting (XSS) from processed content. \n- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill manages authentication credentials in ~/.busabase/airapps. It leverages the busabase-sdk which enforces restrictive owner-only file permissions to protect these secrets. \n- [COMMAND_EXECUTION]: Several local scripts (ingest_plan.mjs, run_checks.mjs, execute_decisions.mjs, export_plans.mjs) execute shell-level operations for data ingestion, auditing, and document export. \n- [UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The project uses verified Node.js packages such as hono and busabase-sdk. A dedicated check script ensures all critical dependencies remain exact-pinned to safe versions.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 25, 2026, 10:36 AM
Security Audit — agent-trust-hub — kelly-lesson