agent-pd checks what Claude Code agents did
A Hacker News post introduced agent-pd. It records tool use and permission events from Claude Code’s main agent and subagents into an audit log. The pd command then reads that log and uses six detectors to report rule offenses. The README says it reports what happened but does not block actions.
Key points
- agent-pd records activity from Claude Code’s main agent and subagents.
- It logs tool calls, permission denials, and subagent start and stop events.
- pd report shows suspected rule offenses for a session.
- pd watch shows new agent activity while work is happening.
- The README says agent-pd does not replace permission prompts or a sandbox.
Quick term guide
- subagents
- Smaller, specialized AI helpers that work under a main AI system to handle specific tasks.
- subagent
- A separate Claude instance that handles one specific task at the same time as other subagents, enabling parallel work.
- audit log
- A record of actions, such as who changed what and when.
- visibility
- How easily people can discover and notice a product.
- sandbox
- A separate space where code can run without affecting other work.
- tool calls
- Times when an AI system uses another function, such as search or file access.
- tool call
- One time an AI agent uses a tool, such as search, calculation, or file reading.
- permission prompt
- A message that asks the user to allow an action before it continues.