Problem

Slop is usually not a mysterious model failure. It is untracked residue: compatibility shims, unmigrated callers, duplicated concepts, stale docs, tests for old contracts, dead paths, and tool gaps that were never represented as work objects.

Current agents can finish a task while leaving that residue invisible. The next agent then rediscovers it, amplifies it, or mistakes it for intended architecture.

Desired behavior

Add a durable SlopLedger work object that any workflow can update.

Ledger buckets:

retained_compatibility
unmigrated_callers
duplicate_concepts
naming_drift
stale_docs
stale_tests
suspected_dead_code
unverified_public_behavior
tool_gaps
accepted_debt

The ledger should be source-linked, queryable, and attached to tasks/handoffs/evidence. It should work for refactors, releases, GitHub triage, docs, provider migrations, and Model Lab traces.

Acceptance criteria

• Define a SlopLedger schema with severity, confidence, owner, source links, status, and cleanup recommendation.
• Add a model/tool workflow for appending ledger items during task completion and refactor review.
• Add a query/read surface for current task, current repo, and handoff export.
• Add a redacted GitHub issue export path for accepted debt items.
• Integrate with verifier/review flows so "done" can be done with accepted residue rather than silently clean.
• Tests cover append/update/resolve/export and stale source handling.

Related

• Refactor completion gate: architecture boundaries, ADRs, and module ownership #1899 refactor completion gate.
• Agent routing: detect when parent work should delegate to scouts or RLM #2024 agent routing.
• RLM/log workbench: route large local logs and structured payloads out of the parent transcript #2023 RLM/log workbench.
• Model Lab: connect CodeWhale traces to open-weight fine-tuning/eval services #1977 Model Lab.