Imported from Coconut-ch1ken/AI4Research (.agents/skills/exp-status/SKILL.md). Install upstream with npx skills add Coconut-ch1ken/AI4Research --skill exp-status. Copyright stays with the author.

$exp-status

This is a Solar AutoSci wrapper skill. Keep Solar as the orchestrator.

Required Execution Path

Do not run native AutoSci repo tools directly. Do not mutate AutoSci's original wiki/, tools/, or runtime/ paths from this skill.

When the user invokes this skill, preserve their arguments and route the request through the Solar Harness runtime, not the current worktree copy:

"${HARNESS_DIR:-$HOME/.solar/harness}/solar-harness.sh" autosci '$exp-status <user args>'

If HARNESS_DIR is unset but solar-harness is on PATH, this explicit AutoSci subcommand form is also valid:

solar-harness autosci '$exp-status <user args>'

Keep the AutoSci subcommand explicit and quote the dollar command in shell contexts so it is not expanded as an environment variable.

Solar Route

• Solar capability: cap.research-experiment-monitor
• Backend action: monitor_experiment
• Coverage status: partial
• Side-effect policy: none

Human-Facing Outputs

After execution, report the Solar-managed run evidence and the human-facing workspace paths:

• Solar-managed run evidence: harness/artifacts/autosci/runs/<run-id>/
• Human-facing wiki: harness/artifacts/autosci/workspace/wiki/
• Human-facing outputs: harness/artifacts/autosci/workspace/wiki/outputs/

Logs, envelopes, retries, and operator state remain Solar-managed and should not be copied into the human-facing wiki.

Limitations

• Local status parity covers wiki/run-state lookup, local session registry state, supplied approved runtime evidence, and collect-ready detection without remote/provider configuration.
• Local run-dir status files and supplied runtime evidence are not sufficient to claim live SSH/provider polling; live remote/provider polling remains optional remote-mode proof.