Reviews run evidence and workflow config to recommend workflow improvements.
npx skills add https://github.com/sandsower/beislid --skill retro-sandsower
Please help me install this Agent Skill. Skill: retro Source: https://github.com/sandsower/beislid/tree/main/skills/retro Command: npx skills add https://github.com/sandsower/beislid --skill retro-sandsower
.beislid/workflow.md and produces actionable, recommendation-only improvement proposals. ## Core Features & Use Cases - Evidence-Based Review: Collects git state, checkpoints, run-ledger summaries, and session context, then compares them against the entire workflow configuration. - Conversational Recommendations: Groups findings into changes worth making, friction to leave alone, decisions needing your call, and non-config notes. - Setup Handoff: Generates a paste-ready handoff block so accepted changes route through setup, the canonical config writer—retro never edits workflow.md itself. - Use Case: After a run where a gate kept firing false positives, invoke retro to analyze the evidence and receive a concrete gate-tuning recommendation plus a setup handoff. ## Quick Start Run a retro on this session and tell me what should change in our Beislið workflow.High-intent search queries and answers about installing and using this skill.
Invoke the retro skill after a run or session. It loads available evidence such as git state, checkpoints, and run-ledger summaries, compares it against .beislid/workflow.md, and presents conversational recommendations for what to change.
Use retro to review skipped, noisy, missing, or slow gates against actual run evidence. It suggests concrete gate-set and selector changes, then produces a paste-ready handoff you route through the setup skill to apply.
No, retro is recommendation-only and never modifies .beislid/workflow.md. The setup skill remains the canonical config writer; retro only builds a setup handoff describing sections to change and recommended defaults.
Retro uses whatever is available: session context, git status, workflow.md, checkpoints, run ledgers, and memory notes. All sources are optional—if evidence is missing or stale, it says so and continues with what it has.
Use doctor when the question is config validity or probe health—whether the setup works. Use retro when the question is what should change after a run, such as refining gates, defaults, or workflow prose based on observed friction.
Leap0920Persist multi-step goal progress across Claude sessions with checkpointed artifacts.
EAIconsultingConvert recurring manual tasks into documented Cowork workflows with run-this-workflow.md and workflow-design.md.
davepoonRenders a compact status panel showing configured integrations as ok, missing, or not-configured.
nguyenpv1980-wqConvert business and technical risks into a layer-based QA test plan.