The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies extensively on shell command execution and script invocation to manage session lifecycles and automate terminal applications.
Evidence: JS and Bash scripts in the scripts/ directory use child_process.execSync, spawnSync, and osascript to interact with system tools and the cmux CLI.
[DATA_EXFILTRATION]: The skill accesses sensitive conversation transcripts and project history stored in the ~/.claude/ directory. While this is the intended purpose of the skill, it involves reading potentially private code and metadata.
Evidence: scripts/search-sessions.js and scripts/list-sessions.js parse JSONL files in the project session directories to extract message content and summaries.
[PROMPT_INJECTION]: The skill displays content from historical session transcripts (JSONL files), which represents an indirect prompt injection attack surface if those transcripts contain malicious instructions.
Ingestion points: ~/.claude/projects/*/*.jsonl (session history transcripts read by search and list tools).
Boundary markers: Absent; transcript content is displayed directly in the terminal output.
Capability inventory: The skill has permissions to execute arbitrary shell commands via the platform's Bash tool and its own scripts.
Sanitization: search-sessions.js includes a NOISE_PATTERNS filter to ignore standard system context files like CLAUDE.md and MEMORY.md during searches.

claude-tracker-suite