The Agent Skills Directory

[DATA_EXPOSURE]: The skill captures and logs sensitive network information, including Authorization headers and browser cookies, as part of its core API tracing functionality. This data is persisted to the system's temporary directory within session files.
[COMMAND_EXECUTION]: The tool manages a background daemon process using child_process.spawn to facilitate continuous network monitoring without blocking the main agent session.
[DYNAMIC_EXECUTION]: The report function generates shell commands (curl) that incorporate untrusted network data, such as URLs and header values, without proper shell escaping. This presents a potential command injection risk if the generated output is directly executed in a terminal.
[DATA_EXFILTRATION]: No evidence of unauthorized data transmission was found. The skill records data for local analysis and report generation, requiring manual action to move or share the captured information.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it ingests untrusted data from network responses and interpolates it into reports without sanitization or explicit boundary markers.
Ingestion points: Captured via CDP Network.loadingFinished and Network.getResponseBody in lib/recorder.js.
Boundary markers: Absent in report templates within lib/reporter.js.
Capability inventory: File system writes for session storage and spawn for daemon management.
Sanitization: No validation or escaping is applied to network data before it is included in the Markdown or cURL reports.

api-tracer