Classify run inputs into four proposal buckets and write a reflection file.
npx skills add https://github.com/lookatitude/guild --skill guild-reflect
Please help me install this Agent Skill. Skill: guild-reflect Source: https://github.com/lookatitude/guild/tree/main/skills/meta/reflect Command: npx skills add https://github.com/lookatitude/guild --skill guild-reflect
Post-task reflection helps teams capture learning from a Guild run, convert outcomes into concrete improvement proposals, and ensure memory is preserved without mutating wiki content or decisions.
Trigger the reflect skill after a successful verify-done to create the reflection file at .guild/reflections/<run-id>.md and begin review.
High-intent search queries and answers about installing and using this skill.
Post-task run insights are captured by classifying observations into four proposal buckets: skill_improvement, missing_specialist, context_issues, and followup_backlog. The process extracts verbatim evidence from run inputs and writes a single reflection file at .guild/reflections/<run-id>.md.
You generate a reflection file only after a successful verify-done phase with at least one relevant action. This gating ensures that retrospective artifacts are created exclusively behind meaningful work, preserving accurate memory without mutating wiki content or decisions.
Observations are extracted by identifying and classifying data from summary.md, handoffs/*.md, and verify.md inputs. The mechanism categorizes these inputs into distinct proposal buckets, producing a concise rationale and including verbatim evidence for each non-empty category.
The reflection process never writes to the wiki or decisions repository. It outputs a single reflection file with frontmatter metadata and emits a handoff to /guild:stats when integrated, leaving content promotion to separate wiki-ingest or decisions components.
Yes, the reflection file specifically tracks context_issues and missing_specialist roles alongside skill_improvement and followup_backlog proposals. This categorization helps teams identify workflow gaps and role deficiencies directly from the run's verbatim evidence.
The best way is to generate a standalone reflection file at .guild/reflections/<run-id>.md. This approach captures learning and preserves team memory by surfacing proposals for review without directly mutating any existing wiki or decisions content.
kyldvsConvert vague project ideas into Claude-executable PLAN.md files with atomic tasks and checkpoints.
GGEduIngest documents to synthesize and update wiki pages in docs/src/wiki.
Embers-of-the-FireGenerate ML experiment code with reproducible PyTorch or scikit-learn workflows.
HeartcoolmanClassify pull requests into changelog categories like NEW-FEATURE and BUG-FIX.
