Skip to content

CLI (asd)

Top-level commands: init, index, read, ledger, think, policy, verify-effects, trace, sync, hydrate, audit, hooks, mcp, conclusions.

Each command is also available via clap aliases that match the MCP tool names, so agents trained on MCP-era docs can type either asd ledger append or asd ledger_append.

Record and retrieve agent reasoning entries. Four entry types:

SubcommandEntry kind
asd think speculate <qname> --conf <0.0–1.0> --summary <text>hypothesis
asd think model <name> --symbols <q1,q2,...> --summary <text>mental_model
asd think failed <qname> --tried <text> --because <text>failed_attempt
asd think question <qname> --q <text>open_question
asd think list [--kind <kind>] [--symbol <qname>] [--json]

Bootstrap helpers:

  • asd think bootstrap — print the initial-read checklist
  • asd think bootstrap --check — scan existing entries and report gaps vs. inherited thinking
  • asd think prompt — print the embedded initial-read prompt template

See Agent Thinking for the full guide.

Export and import thinking entries for cross-session or cross-repo sharing:

  • asd conclusions export — write entries to .asd/conclusions/*.jsonl
  • asd conclusions import — load from .asd/conclusions/ (runs automatically on git pull via post-merge hook)