Skip to content

TODO.md

Latest commit

 

History

History
94 lines (71 loc) · 4.23 KB

TODO.md

File metadata and controls

94 lines (71 loc) · 4.23 KB

TODO

Forward-looking follow-ups and proposal-only backlog for AutoLabOS.

Usage rules:

  • This file is for pending follow-ups, operator UX backlog, and proposal-only work.
  • This file does not replace ISSUES.md.
  • ISSUES.md is reserved for reproduced live-validation defects and tracked research/paper-readiness risks.
  • Canonical contracts still live in AGENTS.md and docs/.

Now

  • Make live-validation temp workspace builders emit run_status.json with validation_scope=live_fixture.

    • Current temp workspaces can still look noisier than intended in /doctor because legacy fixtures do not emit the new scope marker.
    • Keep this additive; do not weaken the default governed-run validator.

  • Decide whether to add an additive backfill helper for legacy runs that predate run_status.json.

    • Current behavior is honest fallback projection.
    • Follow-up is whether a bounded backfill would improve operator surfaces without pretending old runs had artifacts they never wrote.

  • Add a per-run completeness checklist artifact.

    • Summarize presence of run_record.json, events.jsonl, checkpoints, key node artifacts, and public mirrors.
    • Keep this separate from harness validation and do not turn it into a new gate by default.

Next

  • Strengthen VerifiedRegistry beyond metadata verification toward bounded content-equivalence checks.

    • Current state:
      • local corpus diagnosis
      • bounded external metadata lookup
      • repair/reject logging
    • Not yet implemented:
      • DOI landing content checks
      • arXiv page/content comparison
      • stronger content-level equivalence heuristics

  • Promote the ad hoc real browser/TUI validation scripts into repo-owned validation helpers if they become recurring.

    • The repo now has real validation seams, but the live browser scripts used during recent work are still operator-local rather than first-class repo utilities.

  • Add an additive artifact-to-output mapping helper note.

    • operator_summary.md is now clearer, but there is still no single thin helper that maps run-scoped canonical artifacts to public-facing outputs.

Later

  • Add a documented workflow-change proposal template in docs/.

    • Documentation only.
    • It should include:
      • affected node(s)
      • preserved contracts
      • review-gate impact
      • rollback safety

  • Consider a richer operator-history browsing surface in TUI/web if operator_history/* becomes a frequent navigation target.

    • Keep operator_history/* additive and human-readable only.
    • Do not replace canonical JSON artifacts.

Proposal-only

These came up in planning but are not current AutoLabOS contract and should not be treated as implemented behavior.

  • GATE_MODE=human / none / auto style gate control

    • Current repo contract already uses approval_mode and execution_approval_mode.
    • Any GATE_MODE-style surface would be a new policy layer.

  • THEORY_PIPELINE and THEORY_MAX_ITERATIONS

    • Mentioned as useful theory-mode controls.
    • Not part of the current repo contract.

  • OpenClaw-style long-running bridge adapters

    • use_message
    • use_cron
    • use_sessions_spawn
    • use_web_fetch
    • use_browser
    • These stay proposal-only until they are explicitly designed into current operator/runtime contracts.

  • results-analysis and results-report as new canonical artifact families

    • Current repo contract keeps existing per-run JSON artifacts as canonical.
    • Human-readable reporting should remain a thin helper layer unless the artifact contract is intentionally revised.

  • Venue-pattern writing memory with strong trust scoring and automatic reload behavior

    • EpisodeMemory exists, but venue-style memory is not a current write_paper contract.
    • If adopted, it should start as advisory-only and stay clearly separated from hard gates.

  • Community template registry with mandatory schema/routing/reproducibility gating

    • Useful future direction for experiment template intake.
    • Not yet a current top-level AutoLabOS contract.

Not In Current Contract

  • Any new top-level workflow node
  • Any redefinition of review as merely a polish stage
  • Any reporting layer that replaces canonical per-run artifacts
  • Any new orchestration engine that bypasses runs.sqlite / run_record.json / events.jsonl