The Agent Skills Directory

[Data Exposure & Exfiltration] (SAFE): The skill accesses task files located in ~/.claude/daily-tasks/. This is local file access necessary for the skill's primary function. No evidence of data transmission to external servers or unauthorized access to sensitive system files was found.
[Command Execution] (SAFE): Uses the cat and date commands to retrieve file contents and timestamps. These commands are executed with system-provided date strings and do not include unsanitized user input in the shell execution context.
[Indirect Prompt Injection] (LOW): The skill reads from local JSON files that could theoretically contain malicious instructions. However, the logic is highly structured and uses the disable-model-invocation: true flag, which prevents the AI from dynamically interpreting content from the file as new instructions. 1. Ingestion points: ~/.claude/daily-tasks/*.json 2. Boundary markers: Not explicitly implemented. 3. Capability inventory: File reading, writing (via Write tool), and basic system command execution. 4. Sanitization: No explicit sanitization of JSON content, but logic is restricted to predefined fields.

daily-done