Slice 5: record browser tasks with tyto learn

[rvegajr]

Stacked on #8. Spec sentences are in docs/IMPLEMENTATION.md §4 Slice 5.

What

tyto learn <name> [--session s] records what an agent does in an agent-browser session, by listening to agent-browser's own event stream (no wrapper or shim). tyto learn status lists the typed inputs by name and field. tyto learn stop <name> --task "…" [--param input_N=name] saves a trace for the compiler (Slice 6).

• Core
  • actionToArgv maps the stream's internal action names (navigate, getbylabel, waitforurl, …) back to CLI argv. The mapping comes from agent-browser's commands.rs.
  • buildTrace:
    • pairs command and result by id;
    • flags gaps (orphan result, missing result, closed socket) as lossy;
    • replaces typed values with {{input_N}}, or {{param}} when kept;
    • masks password-like fields;
    • reduces cookie and storage values to names and lengths;
    • redacts outputs.
  • Recorder records only after a get url sync marker.
  • New ports BrowserEventSource and TraceStore, with fakes.
• agent-browser adapter: StreamEventSource mirrors agent-browser's socket-dir rules (AGENT_BROWSER_SOCKET_DIR, then XDG_RUNTIME_DIR, then ~/.agent-browser, with namespaces).
• store: FileTraceStore (0600).
• cli:
  • learn start/status/stop;
  • a detached listener that keeps typed values in memory only;
  • control over a unix socket in a 0700 directory.

Found while probing

• The stream's success field is false even for successful commands, so success is derived from the result's error.
• Typed values (passwords included) are broadcast in plaintext, so they're masked at build time and never written unless kept.

Tests

• npm run check: 14 files, 122 tests, typecheck clean, gitleaks clean.
• npm run test:live: 6 passed. The new test records a real session against a loopback page: 3 steps in CLI form, with only the named input kept.
• Manual end to end (Wikipedia search, done through plain agent-browser):
  • 6 steps recorded in CLI form, no gaps;
  • the search term kept as {{species}} (example "Snowy owl");
  • snapshot and eval result ("Vulnerable") kept as evidence for the compiler;
  • trace written 0600; the listener exited on stop.

🤖 Generated with Claude Code