The Agent Skills Directory

Data Exposure (LOW): The skill accesses session logs stored in ~/.dotfiles/claude.symlink/projects/. These files contain full chat histories, which are sensitive. However, this access is directly related to the skill's primary purpose.
Indirect Prompt Injection (LOW): The skill ingests untrusted data from session logs which could contain malicious instructions designed to influence the agent reading the resulting report.
Ingestion points: .jsonl files located in the project subdirectories.
Boundary markers: Absent; data is processed and printed directly to the output.
Capability inventory: The skill uses local bash commands (jq, grep, cat, etc.) to parse and aggregate data.
Sanitization: Absent; the script extracts raw strings from the logs using jq without escaping or filtering.
Command Execution (SAFE): The skill uses a strictly defined list of allowed tools (jq, find, wc, etc.) and performs local data processing only. No network activity or unauthorized system modifications were detected.

context-report