#592 Apply recommendation gradient to all substantive option choices

packages/agents/content/guidance/shared/AGENTS.md

@@ -39,9 +39,9 @@ Capitalize the first word after a colon, unless the colon falls mid-sentence and
 
 When prompting the user for input, use visual markers to make prompts more noticeable:
 
-- **Confirmation prompts** (asking the user to confirm a proposed action): End with `👍🏼👎🏼`.
+- **Confirmation prompts** (the user's response is approve-or-redirect; "no" means "let's adjust or discuss," not a concrete alternative action): End with `👍🏼👎🏼`.
 - **All other questions** (open-ended, clarifications): End with `🤔`
-- **Numbered options (2+ choices)**: Follow the recommendation-gradient convention — mark each option ■■■/■■□/■□□/□□□ and list `➕` pros and `➖` cons. Skip the gradient for confirmation prompts and next-steps menus. When a response contains 2+ option-style questions, prefix each with `Q1`, `Q2`, etc. Full spec: `_data/recommendation-gradient.md` in the agents skills tree.
+- **Numbered options (2 or more choices)**: Follow the recommendation-gradient convention, marking each option ■■■/■■□/■□□/□□□ and listing `➕` pros and `➖` cons. This covers every option-style list with substantive tradeoffs, including templated next-steps menus and yes/no choices where both paths are concrete actions (rendered as a 2-option gradient list rather than `👍🏼👎🏼`). When a response contains 2+ option-style questions, prefix each with `Q1`, `Q2`, etc. Full spec: `_data/recommendation-gradient.md` in the agents skills tree.
 
 Examples:
 

packages/agents/content/skills/_data/complexity-classification.md

@@ -43,12 +43,12 @@ Classify the complexity of a task, finding, or change to determine how it should
 
 Each consuming skill defines which complexity levels qualify for its "simple enough" decision against this rubric.
 
-| Consumer                           | Levels | Decision                                                                                                                                                                                 |
-| ---------------------------------- | ------ | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| `wrap-up` drive-by pass            | 1–2    | Apply immediately on the current branch; skip ticket creation                                                                                                                            |
-| `next-steps-after-plan`            | 1–2    | Recommend "Implement directly with follow-up review". When a review pass would catch nothing meaningful (e.g., a typo or unused-import removal), recommend "Implement directly" instead. |
-| `next-steps-after-review`          | 1–2    | Recommend "implement directly" (findings are simple enough to fix without further review)                                                                                                |
-| `orchestrated-architect` (planned) | 1–2    | Classify as `none`/`low` impact (minimal architectural guidance)                                                                                                                         |
+| Consumer                           | Levels | Decision                                                                                                                                                                                          |
+| ---------------------------------- | ------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `wrap-up` drive-by pass            | 1–2    | Apply immediately on the current branch; skip ticket creation                                                                                                                                     |
+| `next-steps-after-plan`            | 1–2    | Recommend "Implement directly with follow-up review". When a review pass would catch nothing meaningful (e.g., a typo or unused-import removal), recommend "Implement directly" instead.          |
+| `next-steps-after-review`          | 1–2    | Recommend "Implement directly with follow-up review". When a re-review would catch nothing meaningful (e.g., a single typo fix or unused-import removal), recommend "Implement directly" instead. |
+| `orchestrated-architect` (planned) | 1–2    | Classify as `none`/`low` impact (minimal architectural guidance)                                                                                                                                  |
 
 When characteristics span two levels, prefer the higher level. This is consistent with the "when uncertain, recommend the more thorough option" pattern used by consuming skills.
 

packages/agents/content/skills/_data/next-steps-after-plan.md

@@ -4,41 +4,46 @@ Standard next-steps block for skills that produce or refine an implementation pl
 
 ## Options
 
-| #   | Emoji | Option             | Description                                      |
-| --- | ----- | ------------------ | ------------------------------------------------ |
-| 1   | 🧠    | Refine plan        | Review the plan for completeness and correctness |
-| 2   | 🎶    | Orchestrate        | Run the full orchestrated development pipeline   |
-| 3   | 🚀    | Implement directly |                                                  |
+| #   | Emoji | Option                                   | Description                                                             |
+| --- | ----- | ---------------------------------------- | ----------------------------------------------------------------------- |
+| 1   | 🧠    | Refine plan                              | Review the plan for completeness and correctness                        |
+| 2   | 🎶    | Orchestrate                              | Run the full orchestrated development pipeline                          |
+| 3   | 🚀🔍  | Implement directly with follow-up review | Implement, then run a single end-of-work review pass as a separate step |
+| 4   | 🚀    | Implement directly                       | Implement without a follow-up review (reserved for trivial work)        |
 
 ## Output format
 
-Present all three options as a numbered list. The recommendation rules below determine which option is recommended — bold that option's label and append `(🟢 recommended)`. Include all known paths (plan, ticket) in each option line; omit paths that are not available in the current context. Use `~/`-relative paths where possible and absolute paths otherwise.
+Present all four options as a numbered list in [recommendation-gradient](./recommendation-gradient.md) form. Each option carries a strength marker (■■■/■■□/■□□/□□□); the recommendation rules below determine which option earns the strongest marker. Pros and cons are omitted by default — add a `➕` or `➖` line only when the specific plan presents a context-specific tradeoff bearing on which option fits (e.g., "plan introduces a new dependency boundary," "single module with no downstream effects"). Generic option properties ("structured review pass," "longer wall time") are noise and must be omitted; see the [recommendation gradient's don'ts](./recommendation-gradient.md#donts) for the rule. Include all known paths (plan, ticket) in each option line; omit paths that are not available in the current context. Use `~/`-relative paths where possible and absolute paths otherwise.
 
 Options that invoke a skill include context-clearing guidance:
 
-- **Refine plan** and **Orchestrate**: Prepend "Clear context and use..." — the plan artifact is self-contained, and orchestration dispatches fresh subagents, so prior conversation wastes tokens and can introduce bias.
-- **Implement directly**: No "Clear context" prefix — conversation history is valuable for manual implementation.
+- **Refine plan** and **Orchestrate**: Prepend "Clear context and use..." because the plan artifact is self-contained and orchestration dispatches fresh subagents; prior conversation wastes tokens and can introduce bias.
+- **Implement directly with follow-up review** and **Implement directly**: No "Clear context" prefix; conversation history is valuable for manual implementation. The follow-up-review variant adds a separate `review-branch` step after implementation.
 
-Example:
+Example (rendered for the default case, where the recommendation rules below select Orchestrate):
 
 ```
 Next steps:
-  1. 🧠 Refine plan:
-     Clear context and use the `refine-plan` skill with
-     plan: {plan_path},
-     ticket: {ticket_source}
-  2. 🎶 **Orchestrate** (recommended):
-     Clear context and use the `orchestrate-dev` skill with
-     plan: {plan_path},
-     ticket: {ticket_source}
-  3. 🚀 Implement directly
+1. 🧠 ■□□ Refine plan:
+   Clear context and use the `refine-plan` skill with
+   plan: {plan_path},
+   ticket: {ticket_source}
+2. 🎶 ■■□ Orchestrate:
+   Clear context and use the `orchestrate-dev` skill with
+   plan: {plan_path},
+   ticket: {ticket_source}
+3. 🚀🔍 ■□□ Implement directly with follow-up review:
+   Implement directly, then clear context and use the `review-branch` skill with
+   ticket: {ticket_source}
+4. 🚀 ■□□ Implement directly
 ```
 
 Skill names for each option:
 
 - 🧠 **Refine plan** -> `refine-plan`
 - 🎶 **Orchestrate** -> `orchestrate-dev`
-- 🚀 **Implement directly** -> no skill invocation; the user implements manually or asks the agent to begin
+- 🚀🔍 **Implement directly with follow-up review** -> no plan-time skill invocation; implement manually, then run `review-branch` (or `orchestrate-review`) as a separate post-implementation step
+- 🚀 **Implement directly** -> no skill invocation; implement manually or ask the agent to begin
 
 ## Recommendation rules
 
@@ -57,8 +62,13 @@ Select the recommended option by checking these rules in order and stopping at t
    - The plan has not been previously refined
    - A prior iteration of `refine-plan` resulted in significant alteration of the plan or significant expansion of the scope of the changes required to implement the plan
 
-2. **Implement directly** — recommend when the work's verification surface fits a single end-of-work review pass: Single module/package, the plan is precise (or follows an established pattern closely), and the implementation's consequences are bounded enough that compiler + tests + one review pass would catch the meaningful classes of mistake. Cross-cutting changes, novel patterns, or work whose consequences ripple beyond the immediate change site fall through to rule 3.
-3. **Orchestrate** — all other cases (default)
+2. **Implement directly with follow-up review** — recommend when the work's verification surface fits a single end-of-work review pass: single module/package, the plan is precise (or follows an established pattern closely), and the implementation's consequences are bounded enough that compiler + tests + one review pass would catch the meaningful classes of mistake. The default for non-trivial bounded work; trivial work at complexity levels 1–2 falls to rule 3.
+3. **Implement directly** — recommend instead of rule 2 when the work is trivial enough that a review pass would catch nothing meaningful (e.g., a typo fix, unused-import removal, single-file mechanical rename). Complexity levels 1–2 trivial only.
+4. **Orchestrate** — all other cases (default). Cross-cutting changes, novel patterns, or work whose consequences ripple beyond the immediate change site fall here.
+
+### Marker strengths
+
+The selected option carries the ■■□ marker in the rendered output. The other three options carry ■□□ by default. Reserve □□□ for an alternative with a clear drawback in the current context. Reserve ■■■ for the selected option only when you would actively push back against any other choice. See [recommendation-gradient markers](./recommendation-gradient.md#markers) for the full marker table and worked examples of the ■■■ and □□□ cases.
 
 Each skill supplies its own recommendation context (e.g., whether the plan was developed interactively, whether a review just completed). Apply these rules using that context.