end-of-day-review

Warn

Audited by Gen Agent Trust Hub on Aug 5, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses several shell commands to gather activity data, including git log to view recent commits and tail to read the last 200 lines of Claude's usage logs. It also executes a dynamic Python command (python3 -c "...") to process log statistics.
  • [DATA_EXFILTRATION]: The skill accesses sensitive local files located in the ~/.claude/ directory, specifically usage.jsonl and audit.log. These files contain a detailed history of the user's interaction with the AI, including tool calls and potentially sensitive project metadata. Accessing these files represents data exposure of the session history.
  • [PROMPT_INJECTION]: The skill ingests untrusted data from git log and direct user input, which is then used to synthesize the day's summary and determine "tomorrow's hint" for the daily plan. The lack of boundary markers or sanitization for these external inputs creates a surface for indirect prompt injection, where malicious commit messages or log entries could influence the agent's future planning or summaries.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 5, 2026, 04:49 PM
Security Audit — agent-trust-hub — end-of-day-review