session-health
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill accesses local directories containing sensitive session logs and metadata from AI coding tools. Evidence includes reading from paths like ~/.deepcode/projects/ and ~/.cursor/projects/ to extract usage data.
- [COMMAND_EXECUTION]: The skill utilizes shell utilities such as ls, wc, jq, and sqlite3 to query the file system and parse structured data files. These commands are executed locally to identify active sessions and calculate token counts.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests session data which may contain untrusted content from user conversations. 1. Ingestion points: Reads from session-index.json and various JSONL files containing user and assistant turns. 2. Boundary markers: No explicit delimiters or instructions are used to distinguish log data from analysis logic. 3. Capability inventory: Commands are limited to local read-only operations and git status checks. 4. Sanitization: No sanitization is performed on data retrieved from JSON or SQLite fields before interpolation into the health report.
Audit Metadata