Reflections

Classify local CLI agent sessions and turn the patterns into suggestions that raise autonomous loop rate. This skill only does session analysis. It does not score repository readiness.

Assumptions derived from the requesting task: project skill, local JSONL first (Alan DB later), all local agents, suggestions not auto-applied.

When Not to Use

• The user wants the 11-dimension repo audit → /agent-readiness in supa-skills.
• The user wants to plan or implement a product feature.
• The user asks how session parsing works, without running it.

Preconditions

• Python 3 on PATH.
• At least one local session root exists: ~/.codex/sessions, ~/.claude/projects, ~/.cursor/projects/**/agent-transcripts (results in terminals/{shell_id}.txt), ~/.grok/sessions/**/updates.jsonl, ~/.kimi-code/sessions/**/wire.jsonl plus sibling logs/kimi-code.log.

Procedure