@agentcentral/cli 0.1.1 → 0.1.3

package/README.md

@@ -53,6 +53,7 @@ The config lives at `${XDG_CONFIG_HOME:-~/.config}/agentcentral/config.json`, wr
 ```bash
 agentcentral tools list                         # key-scoped tools (json)
 agentcentral tools list --format table          # aligned table
+agentcentral tools list --domain inventory      # filter by server domain
 agentcentral tools describe get_orders          # full descriptor
 agentcentral tools schema get_orders            # raw JSON Schema
 
@@ -62,6 +63,12 @@ agentcentral tools call get_orders @args.json
 echo '{"limit":5}' | agentcentral tools call get_orders -
 ```
 
+A new hosted tool is callable (`tools list`, `describe`, `schema`, `call`) with no CLI release. The one
+caveat is domain filtering: `tools list --domain` and `mcp serve --domain` fall back to the bundled
+`TOOL_DOMAINS` registry from `@agent-central/shared` when the server omits `_meta` domain hints, so a
+brand-new tool will not appear under a domain filter until that registry is updated (or the server emits
+the tool's domain via `_meta`).
+
 Write tools default to a non-mutating **preview**. To submit, pass `--yes` and an `--idempotency-key`:
 
 ```bash
@@ -69,6 +76,25 @@ agentcentral tools call update_keyword_bids @bids.json \
   --yes --idempotency-key "$(uuidgen)"
 ```
 
+### Convenience read commands
+
+Thin wrappers over the same hosted read tools. Each maps friendly flags onto the exact parameters its
+backing tool declares. They return facts only and add no recommendations or agentcentral-authored next
+actions.
+
+```bash
+agentcentral inventory health                  # -> get_inventory_health (--limit)
+agentcentral inventory risk --asin B00...       # -> get_inventory_risk_triage (--asin, --limit)
+agentcentral ads performance --since 30d        # -> get_campaign_performance (--since, --limit)
+agentcentral catalog quality --asin B00...      # -> get_listing_quality (--asin, --limit)
+agentcentral actions history --since 7d         # -> get_action_history (--since, --limit; needs agent_actions:read scope)
+agentcentral today --fulfillment-channel AFN    # -> get_todays_sales (current-day only; --fulfillment-channel AFN|MFN, --limit)
+agentcentral orders --since 30d --asin B00...   # -> get_orders (--since, --asin, --fulfillment-channel AFN|MFN, --limit)
+```
+
+`--since` accepts a relative lookback (`30d`) or an absolute `YYYY-MM-DD` date. `today` is current-day
+only and takes no `--since`.
+
 ### Run the stdio MCP bridge
 
 `mcp serve` relays newline-delimited JSON-RPC between a local MCP client and the hosted server. stdout
@@ -92,9 +118,13 @@ agentcentral doctor --format markdown
 
 ## Global flags
 
-Every command accepts: `--format/-f json|ndjson|table`, `--output/-o <path>`, `--profile/-p <name>`,
-`--timeout <ms>`, `--limit <n>`, `--quiet/-q`, `--no-color`, `--verbose`, `--request-id <id>`,
-`--yes/-y`, `--idempotency-key <key>`, `--api-key-stdin`.
+Every command accepts: `--format/-f json|ndjson|table|csv|markdown`, `--output/-o <path>`,
+`--profile/-p <name>`, `--timeout <ms>`, `--limit <n>`, `--quiet/-q`, `--no-color`, `--verbose`,
+`--request-id <id>`, `--yes/-y`, `--idempotency-key <key>`, `--api-key-stdin`.
+
+When `--limit <n>` is set and the result comes back with exactly `n` rows, the CLI emits a best-effort
+truncation WARNING to stderr. The hosted server exposes no truncation flag or total-row count, so this
+is a heuristic, it only fires when `--limit` is set, and it never touches stdout.
 
 ## Exit codes