Skip to main content

Agent Files Report

The Agent Files Report provides a browser-readable view of the Agent's memory and Skills files. It extracts references between files and lets you navigate from a file to the files it references.

Report contents​

The generator recognizes these reference identifiers:

  • ep_... for episodes;
  • proc_... for procedures;
  • app_... for application records; and
  • fail_... for failures.

Each matching file shows the reference type, identifier, and resolved path. Selecting Jump opens the target file, scrolls it into view, and highlights it briefly.

Generate and view the report​

On a device, run:

ssh root@<DEVICE_IP>
cd /userdata/agent_tools
./view_agent_files.sh

The generated report is /userdata/agent/files_report.html. Either copy it to a local machine:

scp root@<DEVICE_IP>:/userdata/agent/files_report.html ~/Desktop/
open ~/Desktop/files_report.html

or stream it over SSH:

ssh root@<DEVICE_IP> "cat /userdata/agent/files_report.html" > report.html

The Agent Web service also exposes:

  • GET /user_files to view the current report; and
  • POST /user_files/regenerate to start asynchronous regeneration. A successful response means regeneration has started; it does not mean that the new report is ready. Poll or reload GET /user_files after the generator finishes.

These routes are served by the Agent Web endpoint on port 8080.

Local implementation​

The report tooling is implemented by:

  • scripts/generate_agent_files_report.py;
  • scripts/agent_files_template.html;
  • scripts/view_agent_files.sh; and
  • src/agent/internal/agent/user_files.go.

The source data is normally under /userdata/agent/memory/ and /userdata/agent/skills/. The report is a generated view and can be regenerated after files change; it is not itself a source of truth.