fix(pipeline): enforce zero-interaction contract across entire go pipeline (v0.1.37)

- flow-code-run: add explicit ZERO-INTERACTION CONTRACT header — no
  AskUserQuestion anywhere in the pipeline
- brainstorm: auto-detect pipeline invocation and force Auto mode
  (both skills/ and codex/skills/ copies)
- plan: "If empty, ask" now conditional on standalone invocation only
- work: "If no input, ask" now conditional on standalone invocation only
- brainstorm in run SKILL.md: explicitly marked as AUTO mode

All sub-skills already had "Never use AskUserQuestion" in review loops.
This closes the remaining gaps in brainstorm and plan/work input handling.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: z23cc

codex/skills/flow-code-brainstorm/SKILL.md

@@ -9,6 +9,8 @@ user-invocable: false
 
 Explore and pressure-test an idea before committing to a plan. Outputs a requirements doc that feeds directly into `/flow-code:plan`.
 
+**Pipeline auto-detection**: If this skill is invoked as part of the `/flow-code:go` pipeline (detected by: epic already exists, or `flow-code-run` is the caller), ALWAYS use Auto mode — AI self-interview with zero human input. Do NOT use `AskUserQuestion`. The go pipeline has a zero-interaction contract.
+
 **IMPORTANT**: This plugin uses `.flow/` for ALL task tracking. Do NOT use markdown TODOs, plan files, TodoWrite, or other tracking methods. All task state must be read and written via `flowctl`.
 
 **CRITICAL: flowctl is BUNDLED — NOT installed globally.** `which flowctl` will fail (expected). Always use:

codex/skills/flow-code-plan/SKILL.md

@@ -72,7 +72,8 @@ Examples:
 - `/flow-code:plan fn-1` (legacy formats fn-1, fn-1-xxx still supported)
 - `/flow-code:plan fn-1-add-oauth then review via /flow-code:plan-review`
 
-If empty, ask: "What should I plan? Give me the feature or bug in 1-5 sentences."
+If empty and invoked standalone: ask "What should I plan? Give me the feature or bug in 1-5 sentences."
+If empty and invoked from `/flow-code:go` pipeline: this should never happen (go always provides input). If it does, derive the plan target from the epic title or requirements doc — never block on user input.
 
 ## Context Analysis (replaces setup questions)
 

codex/skills/flow-code-work/SKILL.md

@@ -56,7 +56,8 @@ Examples:
 - `/flow-code:work Add rate limiting`
 - `/flow-code:work fn-1-add-oauth then review via /flow-code:impl-review`
 
-If no input provided, ask for it.
+If no input provided and invoked standalone: ask for it.
+If invoked from `/flow-code:go` pipeline: input is always the epic ID from the plan phase — never block on user input.
 
 ## Context Analysis (replaces setup questions)
 

skills/flow-code-brainstorm/SKILL.md

@@ -10,8 +10,10 @@ user-invocable: false
 Explore and pressure-test an idea before committing to a plan. Outputs a requirements doc that feeds directly into `/flow-code:plan`.
 
 **Two modes:**
-- **Interactive** (default): asks user questions via `AskUserQuestion`
-- **Auto** (`--auto`): AI self-interview — analyzes codebase, asks itself questions, answers from code evidence, produces refined spec with zero human input
+- **Interactive** (default when invoked standalone via `/flow-code:brainstorm`): asks user questions via `AskUserQuestion`
+- **Auto** (`--auto`, or when invoked from `/flow-code:go` pipeline): AI self-interview — analyzes codebase, asks itself questions, answers from code evidence, produces refined spec with zero human input
+
+**Pipeline auto-detection**: If this skill is invoked as part of the `/flow-code:go` pipeline (detected by: epic already exists, or `flow-code-run` is the caller), ALWAYS use Auto mode regardless of flags. The go pipeline has a zero-interaction contract.
 
 **IMPORTANT**: This plugin uses `.flow/` for ALL task tracking. Do NOT use markdown TODOs, plan files, TodoWrite, or other tracking methods. All task state must be read and written via `flowctl`.
 

skills/flow-code-run/SKILL.md

@@ -12,6 +12,8 @@ user-invocable: false
 
 Internal pipeline engine. Drives the entire development lifecycle (brainstorm, plan, work, review, close) through flowctl phase next/done. Invoked by `/flow-code:go`.
 
+**ZERO-INTERACTION CONTRACT**: This pipeline runs fully autonomously. You MUST NOT use `AskUserQuestion` at any point. All decisions are made automatically based on codebase analysis, config, and best-practice defaults. If information is missing, use the best available default — never block on user input.
+
 **CRITICAL: flowctl is BUNDLED.** Always use:
 ```bash
 FLOWCTL="$HOME/.flow/bin/flowctl"
@@ -65,7 +67,8 @@ Detect input type to decide whether to execute or skip:
 1. Run: `$FLOWCTL phase done --epic $EPIC_ID --phase brainstorm --json`
 2. Proceed to Plan phase
 
-**Execute brainstorm** (input is natural language — a new idea):
+**Execute brainstorm** (input is natural language — a new idea).
+**This is always AUTO mode — AI self-interview, no AskUserQuestion, no human input.**
 1. **Codebase context**: Search for files related to the request (Grep/Glob for key terms), read git log for recent changes, check existing `.flow/` specs for related work
 2. **Classify complexity**: Trivial (1-2 files) / Medium (clear feature) / Large (cross-cutting)
 3. **Self-interview**: Ask and answer 6-10 Q&A pairs grounded in code evidence. Core questions: