@heyamiko/amiko-cli 0.7.9 → 0.7.10

package/README.md

@@ -1,4 +1,4 @@
-# @heyamiko/amiko-cli (v0.7.9)
+# @heyamiko/amiko-cli (v0.7.10)
 
 Manage wallets, credits, swaps, and MPP marketplace services from the terminal. Works for both human users and AI agents running on OpenClaw.
 
@@ -189,6 +189,25 @@ amiko service call POST /v1/music/plan '{"prompt":"Epic orchestral piece"}'
 
 Audio and image endpoints upload to Supabase Storage and return permanent URLs.
 
+## Confirmation Gate for Paid Commands
+
+Every paid command (`image`, `search`, `amazon search`, `service tts`, `service stt`, `service chat`, `service call`, `credits topup`) requires explicit approval before spending:
+
+- **Human, interactive shell** — the command prints the cost and prompts `Proceed? [y/N]`. Answer `y` to run.
+- **Agent / non-interactive shell** — the command *refuses* to run without `--yes`, with a message telling the agent what to do:
+  
+  ```
+  Refusing to run without --yes (non-interactive shell detected).
+  
+  This command will cost 5 AMIKO.
+  Action: Generate image: a sunset over mountains
+  
+  Quote the cost to the user, get their explicit approval, then re-run with --yes:
+    amiko image "a sunset over mountains" --yes
+  ```
+
+This makes cost confirmation effectively mandatory regardless of which model drives the agent — weak models can no longer silently burn tokens by ignoring the skill docs.
+
 ## Error Output
 
 When a paid command fails (payment already sent, but the upstream service returned 4xx/5xx), the CLI prints the actual upstream response body so you can see the real failure reason:

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@heyamiko/amiko-cli",
-  "version": "0.7.9",
+  "version": "0.7.10",
   "description": "Amiko CLI — swap tokens, manage credits, bridge cross-chain, and call marketplace agents",
   "type": "module",
   "bin": {

package/skills/SKILL.md

@@ -15,8 +15,8 @@ The `amiko` CLI is installed globally. Auth and payments are automatic via `.ami
 
 ## Critical Rules
 
-1. **Always confirm the price with the user BEFORE running any paid command.** This is non-negotiable. For every one of: `topup`, `search`, `image`, `amazon search`, `service call`, `service tts`, `swap send` — you must first quote the exact cost in the relevant unit (AMIKO, $USD, or gas), wait for the user's explicit "yes" / "go ahead" / "do it", and only then run the command. Do not pre-empt; do not batch multiple paid calls; do not use `--yes` unless the user has just approved this specific call.
-2. **After every paid command, report the remaining balance to the user.** The CLI now prints a `Balance: N credits (~$X.XX)` line at the end of every paid command. Include that figure in your reply so the user knows what they have left.
+1. **Every paid command is hard-gated on `--yes` in a non-interactive shell.** The CLI will *refuse* to run `image`, `search`, `amazon search`, `service call`, `service tts`, `service stt`, `service chat`, or `credits topup` unless you pass `--yes`. The refusal is loud: it prints the cost and the command you should re-run with `--yes` appended. This means: **you must first quote the cost to the user, get their explicit approval, and only then append `--yes`**. If the user didn't approve, do not add `--yes`.
+2. **After every paid command, report the remaining balance to the user.** The CLI prints a `Balance: N credits (~$X.XX)` line at the end of every paid command — include that figure in your reply so the user knows what they have left.
 3. **Never retry a failed command.** If a command returns an error, report it and stop. Every paid call costs tokens even on failure.
 4. **Never suggest `amiko login` or `amiko connect`.** These commands don't exist. Auth is automatic via `.amiko.json`.
 5. **Check balance before expensive operations.** Run `amiko credits balance` first if unsure about funds.