On-demand help

asd help is on-demand instruction disclosure: it returns the docs for one feature — synopsis, syntax, params, examples, and gotchas — so an agent can pull exactly what it needs, when it needs it, instead of carrying every tool’s full instructions in context every turn.

asd help                 # full feature catalog
asd help impact          # one feature (a name, or a phrase like "blast radius")
asd help --agent         # machine-readable JSON
asd help --manifest      # this binary's feature index

Version-pinned and identical across surfaces

The docs are compiled into the running binary, so they always match the code — no drift between what asd help says and what the command does. The CLI (asd help) and the MCP help tool return byte-identical payloads, so an agent gets the same answer whether it shells out or calls the tool.

Cross-tool resolution

When ASD and CTXone are both installed, help is backed by a shared cross-tool registry: an unknown topic is resolved by the owning tool (asd ↔ ctx). Ask asd help about a CTXone feature and the response tells you so, and vice versa. asd help --publish writes this binary’s manifest into the shared index so a unified help can discover ASD’s features alongside CTXone’s — the index is tool-keyed, so publishing one never clobbers the other’s entry.

Why it matters for agents

Loading a big always-on instruction block spends tokens on capabilities the agent isn’t using this turn. asd help <topic> costs a couple hundred tokens and returns just the feature at hand — cheaper, and always current. Point your agent at asd help (the asd skill always-on block already does) and let it fetch syntax on demand.