Extract durable, codifiable learnings from iterations with quality-gated triage.
npx skills add https://github.com/eumemic/dev-skills --skill retro-eumemic
Please help me install this Agent Skill. Skill: retro Source: https://github.com/eumemic/dev-skills/tree/main/skills/retro Command: npx skills add https://github.com/eumemic/dev-skills --skill retro-eumemic
The retro skill prevents wasted cycles by turning experience from an iteration, full session, or pause point into durable, codifiable learnings—while avoiding noisy or one-off changes that would bias future agent behavior.
--scope=iteration (loop-driver iteration), --scope=session (full conversation), and --scope=pause (self-invoked when waiting on async work).Use retro to reflect on this iteration by running: /retro --scope=iteration
High-intent search queries and answers about installing and using this skill.
To extract durable learnings from a workflow session, run a retrospective with scoped event analysis to identify recurring, codifiable findings. This process reviews artifacts and classifies high-leverage insights into autonomous memory updates or approval-needed skill edits.
A quality gate in a retrospective filters out noisy, one-off changes to prevent biased future agent behavior. It ensures only recurring, codifiable, and high-leverage findings clear the bar to produce durable workflow improvements.
Run a retrospective after a loop-driver iteration by executing the command with the iteration scope. This triggers candidate-finding triage to evaluate events, applying recurring and codifiable checks to surface actionable dev-infrastructure learnings.
Yes, you can trigger a retrospective automatically when an agent pauses for async work by using the pause scope. This self-invoked analysis evaluates the waiting period to find codifiable workflow improvements without manual intervention.
Action items during a workflow retrospective are classified by routing them into autonomous execution or approval-needed surface actions. Selected types include memory entries, skill edits, scripts, and constellation issues, governed by clear autonomy zones.
DKSangGenerate evidence-grounded retrospective reports for DES data engineering implementation cycles.
HmbownApply a uniform corrective action across a horizontal row of related items.
mvillmowAutomate GitHub issue implementation from branch creation to pull request.
AngelP17Document decisions and learnings across planning, work, review, and compound loops.