The Agent Skills Directory

[COMMAND_EXECUTION]: The skill spawns background Node.js processes to manage the log collector lifecycle.
Evidence: scripts/log-server.mjs uses child_process.spawn to start a detached instance of itself as a background server.
Evidence: scripts/start-collector.mjs acts as a wrapper to spawn the logging service with specific configurations.
[INDIRECT_PROMPT_INJECTION]: The agent is instructed to read and analyze logs that may contain data from untrusted external sources, creating an attack surface where malicious input could influence agent behavior.
Ingestion points: SKILL.md (Step 6) instructs the agent to read complete records from LOG_FILE for analysis.
Boundary markers: Absent. There are no instructions or delimiters provided to ensure the agent ignores potential instructions embedded within the trace logs.
Capability inventory: The skill has the ability to spawn subprocesses (scripts/log-server.mjs), write to the filesystem (Step 4: Instrument the Path), and start a network listener (scripts/collector-server.mjs).
Sanitization: Absent. The workflow does not specify any sanitization or filtering of log content before analysis beyond a basic JSON check.
[DATA_EXFILTRATION]: The skill instruments source code to send runtime state and variables to a network endpoint.
Evidence: assets/browser-log-helper.js and assets/node-log-helper.js facilitate sending data to a LOG_SERVER_URL.
Mitigation: The implementation strictly enforces loopback-only communication (127.0.0.1, localhost) in scripts/collector-server.mjs and scripts/log-server.mjs, preventing data from leaving the local machine under normal operation.

trace-mode