Orchestrate durable work into evidence-backed execution loops with checkpointed progress.
npx skills add https://github.com/islee23520/lfg --skill ulw-loop-islee23520
Please help me install this Agent Skill. Skill: ulw-loop Source: https://github.com/islee23520/lfg/tree/main/skills/ulw-loop Command: npx skills add https://github.com/islee23520/lfg --skill ulw-loop-islee23520
The skill helps you run durable, checkpointed work without losing state, proof, or cleanup discipline across long tasks and context loss.
Use the ulw-loop skill to turn this request into a checkpointed, evidence-backed execution plan and begin with the required bootstrap steps.
High-intent search queries and answers about installing and using this skill.
Checkpoint long-running coding tasks by breaking work into evidence-backed execution loops with append-only ledgers and scoped subagent coordination. This records durable progress and observable evidence before completion.
Recover task state after session compaction or interruption using checkpointed execution loops with append-only ledgers. This restores durable work progress and resumes staged delivery without losing prior context.
Evidence tracking for manual quality assurance works by requiring observable evidence and cleanup receipts in an append-only ledger before task completion is recorded. This ensures reviewable operational work.
Task orchestration with scoped subagents supports staged delivery and worker delegation by coordinating durable work across execution loops. Subagents execute scoped tasks while ledgered progress maintains state.
Checkpointed task orchestration requires cleanup receipts before completion is recorded. This ensures durable work maintains cleanup discipline and leaves no orphaned state across long-running delivery loops.
Avoid checkpointed execution loops for short-lived or trivial software engineering tasks that do not require manual quality assurance, worker delegation, or recovery from context loss. Overhead outweighs benefits without durable work.
AGM1968Define 3-5 binary done criteria and non-goals for weave workflow nodes.
smartmarblesAudit chat logs and PROBE reports to reconstruct agent actions and generate verdict reports.
Leap0920Persist multi-step goal progress across Claude sessions with checkpointed artifacts.
Chang-Shih-YungSave git state, decisions, and pending tasks into a reusable checkpoint.
