seq

Overview

Mine ~/.codex/sessions/ JSONL and ~/.codex/memories/ files quickly and consistently with the Zig seq CLI. Explicit $seq requests and artifact-forensics questions are seq-first. Focus on skill usage, format compliance, token counts, tool-call/tooling forensics, memory-file mining, and session-backed proof.

Trigger Cues

• Explicit $seq asks.
• Questions that ask to extract or query finalized $plan artifacts (<proposed_plan> blocks) for a repo, session, or time window.
• Questions that ask to verify prior session output using artifacts ("use $seq to find it" / "what did that session actually say").
• Artifact-forensics questions about provenance, trace proof, or "show me the artifact that proves it".
• Questions about Codex memory artifacts or provenance ("what's in MEMORY.md", "which rollout produced this memory", "is this memory stale", "how do memory_summary.md and rollout_summaries relate").
• Execution ledger forensics from session traces (Orchestration Ledger, subagent lifecycle events, concurrency counts).
• Concurrency math validation (max_concurrency, effective fanout, occurrences of peak parallelism, planned rows vs actual parallelism).