add observation skill container

[paul-nechifor]

Closes DIM-1505

Problem

• Observation skill is broken when running dimsim simulator
• We have too many independent observation skills.

Solution

• Unify all observation skills into one.
• This way, it no longer goes through each sim, just listen to /color_image

[greptile-apps]

Greptile Summary

This change replaces robot-specific cached-image observation with a shared stream-backed observation skill and wires it into the affected robot agents. A focused runtime check confirmed that ObserveSkill reports transport and decoding failures as camera-frame timeouts, which can lead users to diagnose an inactive camera instead of the underlying stream failure.

Confidence Score: 4/5

Not safe to merge until non-timeout image-stream failures are distinguished from genuine frame timeouts.

A focused executable harness injected a concrete stream failure and observed the shared observation skill convert it into a timeout response, directly demonstrating the incorrect failure behavior.

Files Needing Attention: dimos/agents/skills/observe_skill.py needs exception handling that catches only the stream timeout condition and preserves or separately reports other stream errors.

T-Rex Logs

What T-Rex did

• T-Rex produced a proof for a posted P1 finding and linked it to the corresponding review comment.
• The contract-validation work showed that before capture, color_image.get_next(timeout=5.0) raised RuntimeError, and after capture, skill.observe() returned a timeout result with No camera frame received.
• No production code changes were made; the reviewer recommended narrowing the handler to the actual timeout exception and adding this regression case to the permanent test suite.

_{Ran code and verified through T-Rex}

Comments Outside Diff (1)

1. General comment

   ObserveSkill reports non-timeout stream failures as camera timeouts

   • Bug
     • ObserveSkill.observe() returns EXECUTION_TIMEOUT when color_image.get_next() raises a non-timeout RuntimeError, thereby presenting a transport/decoding failure as a missing camera frame.
   • Cause
     • The broad except Exception at dimos/agents/skills/observe_skill.py:37 catches every stream exception and unconditionally builds the timeout result at lines 38-41.
   • Fix
     • Catch only the concrete timeout exception emitted by the stream implementation; allow transport/decoding exceptions to propagate or map them to a distinct execution-failure result. Add the injected RuntimeError regression test.

   _{Ran code and verified through T-Rex}

_{Reviews (1): Last reviewed commit: "add observation skill container" | Re-trigger Greptile}

[greptile-apps]

Stream failures are mislabeled as timeouts

The broad except Exception converts transport, decoding, and other image-stream failures into EXECUTION_TIMEOUT and tells the caller that no camera frame arrived. A reproduced RuntimeError from color_image.get_next() takes this path, so agents and operators cannot distinguish a broken stream from an inactive camera. Catch only the concrete timeout exception, and propagate or return a distinct execution failure for other errors.

Artifacts

Focused non-timeout ObserveSkill harness source

• The authored harness injects a RuntimeError through the color-image stream and compares direct stream behavior with ObserveSkill behavior, proving the masking path.

Direct stream RuntimeError capture

• The direct color-image stream call raised the injected RuntimeError and exited successfully, establishing the non-timeout input condition.

ObserveSkill masks RuntimeError as timeout capture

• Calling ObserveSkill after the identical RuntimeError injection returned EXECUTION_TIMEOUT and a no-camera-frame message, confirming error masking.

_{Ran code and verified through T-Rex}

[codecov]

❌ 1 Tests Failed:

Tests completed	Failed	Passed	Skipped
4507	1	4506	69

View the top 1 failed test(s) by shortest run time

dimos.cli.test_cli_startup::test_help_startup_time

Stack Traces | 8.23s run time

def test_help_startup_time() -> None:
        """`dimos --help` must finish in under {HELP_TIMEOUT_SECONDS}s."""
        start = time.monotonic()
        result = subprocess.run(
            [sys.executable, "-m", "dimos.cli.dimos", "--help"],
            capture_output=True,
            text=True,
            timeout=HELP_TIMEOUT_SECONDS + 5,  # hard kill safety margin
        )
        elapsed = time.monotonic() - start
        assert result.returncode == 0, f"dimos --help failed:\n{result.stderr}"
>       assert elapsed < HELP_TIMEOUT_SECONDS, (
            f"dimos --help took {elapsed:.1f}s (limit: {HELP_TIMEOUT_SECONDS}s). "
            f"Check for heavy imports in the CLI entrypoint or GlobalConfig."
        )
E       AssertionError: dimos --help took 8.2s (limit: 8s). Check for heavy imports in the CLI entrypoint or GlobalConfig.
E       assert 8.230978411000024 < 8

elapsed    = 8.230978411000024
result     = CompletedProcess(args=['.../dimos/dimos/.venv/bin/python', '-m', 'dimos.cli.dimos', '--help'], returncod...x1b[0m\n\x1b[2m╰──────────────────────────────────────────────────────────────────────────────╯\x1b[0m\n\n", stderr='')
start      = 506.307514042

dimos/cli/test_cli_startup.py:75: AssertionError

To view more test analytics, go to the Test Analytics Dashboard
_{📋 Got 3 mins? Take this short survey to help us improve Test Analytics.}

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
dimensional	🟢 Ready	View Preview	Aug 21, 2026, 6:08 PM

💡 Tip: Enable Workflows to automatically generate PRs for you.