The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to perform a global installation of the claude-code-lens package from the NPM registry using npm install -g claude-code-lens if the tool is not found.
[COMMAND_EXECUTION]: The skill relies on executing multiple shell commands via the cclens CLI tool, including cclens trace list, show, and export, to interact with system trace data.
[DATA_EXFILTRATION]: The skill accesses sensitive information by reading Claude Code session traces. These files, typically stored in ~/.claude-code-lens/, contain internal model reasoning, tool execution logs, and full conversation histories from previous AI interactions.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The agent reads exported Markdown trace files which contain the complete record of previous Claude Code sessions.
Boundary markers: The instructions lack explicit boundary markers or directives to ignore instructions that might be embedded within the trace data being analyzed.
Capability inventory: The skill uses shell command execution (cclens trace) and file reading tools.
Sanitization: There is no evidence of sanitization or validation of the trace content before it is read by the agent for analysis, creating a surface for instructions from previous sessions to influence current agent behavior.

cclens-trace-analyzer