A checkable decision trail for long agent runs.
Maintains an evidence-linked decision log so a reviewer can check what happened during extended agent work.
When to use it
Use for explicit invocation, autonomous or multi-phase runs, or work reviewed after a human steps away.
It changes how the agent documents project work by keeping a local decision log and checking its accuracy before handing back.
What you provide
Routes selection of the review subagent's reasoning level and speed to its Subagents section; whether installation is required is unstated.
This skill
agent-transcripts/
Reads your current Cursor transcript
~/.claude/projects/<slug>/
Reads your current Claude Code transcript
~/.pi/agent/sessions/--<slug>--/
Reads your current Pi transcript