@victortomaili/skill-cli 0.7.0 → 0.7.2

package/CHANGELOG.md

@@ -11,6 +11,28 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Cursor adapter (`.cursor/rules` format) for `init -g` bootstrap.
 - Per-agent hook adapters for automatic `/X` triggering (push model).
 
+## [0.7.2] - 2026-07-08
+
+### Changed
+- Hardened the PROPOSE gate: the proposal turn must contain ONLY the questions
+  (enable + parameters) and then END the turn — NO exploration, NO tool calls, NO
+  task work. The agent must not start the task until the user answers. Previously
+  the wording said "stop and wait" but allowed the agent to keep exploring; now
+  "END YOUR TURN" is explicit in both the AGENTS.md block and the `skill active`
+  footer. Also added "I'll start exploring while I wait" to the forbidden-excuses
+  list.
+
+## [0.7.1] - 2026-07-08
+
+### Added
+- PARAMETERS (2nd rule, per user direction): when a proposed skill lists activation
+  options in its description (a level, language, mode, format, or strictness), the
+  agent must ask the user to choose them in the SAME proposal and the gate stays up
+  until BOTH the enable decision AND the parameters are given. Don't pick parameters
+  for the user; don't start the task until both are settled. Examples are fully
+  generic (no skill-specific names). Reflected in the AGENTS.md block and the
+  `skill active` footer (new PARAMS row).
+
 ## [0.7.0] - 2026-07-08
 
 ### Changed

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@victortomaili/skill-cli",
-  "version": "0.7.0",
+  "version": "0.7.2",
   "description": "Cross-agent skill manager — one global store, clean agent folders, activation via skill.config",
   "license": "MIT",
   "author": "Victor Tomaili <victor@tomaili.com>",

package/src/commands/defaults.js

@@ -43,14 +43,18 @@ export function cmdActive() {
   console.log(c.bold('→ For EACH skill above, decide in your reply:'))
   console.log(c.gray('    LOAD    moves CORRECTNESS/QUALITY → run ') + c.cyan('skill cat <name>') + c.gray(' now.'))
   console.log(c.gray('    PROPOSE moves COST, STYLE, or SPEED → ASK "enable <name>?" about ALL of them,'))
-  console.log(c.gray('            then STOP and WAIT for the answer. Do NOT start the task until they answer.'))
-  console.log(c.gray('            The wait is the POINT — a trade-off is the USER') + c.gray("'s decision; it moves an axis") )
-  console.log(c.gray('            on EVERY reply, so NEVER skip and NEVER say "not relevant here".'))
+  console.log(c.gray('            then END YOUR TURN and WAIT. That turn contains ONLY the questions — NO exploration,'))
+  console.log(c.gray('            NO tool calls, NO task work. Do NOT start the task until they answer. The wait is'))
+  console.log(c.gray('            the POINT — a trade-off is the USER') + c.gray("'s decision; it moves an axis on EVERY") )
+  console.log(c.gray('            reply, so NEVER skip and NEVER say "not relevant here".'))
   console.log(c.gray('            Mentioning / noting / "available" is NOT asking.'))
+  console.log(c.gray('    PARAMS  if a proposed skill lists activation options (level / language / mode /'), )
+  console.log(c.gray('            format / strictness), ask the user to choose them in the SAME proposal — do not pick'))
+  console.log(c.gray('            for them. Gate stays up until enable decision + parameters are both given.'))
   console.log(c.gray('    SKIP    moves NONE of the above.'))
-  console.log(c.yellow('  ⚠ Breaking the gate: starting the task before the user answers,'))
+  console.log(c.yellow('  ⚠ Breaking the gate: starting the task / exploring before the user answers,'))
   console.log(c.yellow('    "would delay it", "user has a clear request", "I will just do the task",'))
-  console.log(c.yellow('    "not relevant here". → ASK first, then wait.'))
+  console.log(c.yellow('    "not relevant here". → ASK first, then END your turn.'))
 }
 
 // `skill default <name>` — mark a skill as a default (active by default in every

package/src/lib/agents-md.js

@@ -38,12 +38,21 @@ no discretion and no third option:
   is you STEALING the user's decision. Deciding a trade-off is not your role;
   surfacing it is. A line in a table is NOT proposing — you must actually ask.
   PROPOSE IS A GATE: ASK "Enable <name>?" (ask about ALL such skills in ONE message),
-  then STOP and WAIT for the user's answer. Do NOT start the actual task until they
-  answer (yes → apply it, no → proceed without). The one-turn wait is the POINT —
+  then END YOUR TURN and WAIT — do nothing else that turn. The proposal turn contains
+  ONLY the questions (enable + any parameters): NO exploration, NO tool calls, NO
+  reading files, NO starting the task. Do NOT begin the actual task until the user
+  answers (yes → apply it, no → proceed without). The one-turn wait is the POINT —
   the user chooses the trade-off BEFORE you start — so "asking would delay the task",
-  "the user has a clear request", "I'll just do the task first", and "I'll mention it
-  and proceed" are FORBIDDEN: each is you deciding for the user. Mentioning / noting /
-  "available" is NOT asking — you must ASK a yes/no question and then wait.
+  "the user has a clear request", "I'll just do the task first", "I'll mention it and
+  proceed", and "I'll start exploring while I wait" are FORBIDDEN: each is you
+  deciding for the user. Mentioning / noting / "available" is NOT asking — you must
+  ASK a yes/no question and then END your turn.
+  PARAMETERS (2nd rule): if the skill's description lists activation options to pick
+  from — e.g. a level, a language, a mode, a format, or a strictness — you MUST ask
+  the user to choose them in the SAME proposal (do NOT pick for them), and the gate
+  stays up until they give BOTH the enable decision AND the parameters. Genericize
+  — never assume a particular skill; read its description to see which options (if
+  any) it exposes, and ask for those.
 - moves none of the above → skip.
 When unsure if a skill moves an axis → PROPOSE (ask). LOADED \u2260 LISTED: a skill is
 loaded only if you \`cat\`-ed it this session — listing it, its \u2605, or its \`active\`