@codebehind/agent-workflow 1.1.14 → 1.1.16

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@codebehind/agent-workflow",
-  "version": "1.1.14",
+  "version": "1.1.16",
   "description": "Scaffold the agent-workflow spec-driven delivery framework into any repo",
   "type": "module",
   "bin": {

package/templates/.agent-workflow/README.md

@@ -94,8 +94,8 @@ When writing final documentation, choose update/create in this order:
 To reduce onboarding cost, use a reusable codebase snapshot file:
 
 - Path: `.agent-workflow/codebase.md`
-- If the file is missing, the first onboarded agent scans the codebase and creates it.
-- If the file exists, agents read it instead of rescanning the codebase.
+- **If the file is missing, create it immediately — before doing any other work.** Scan the codebase and write the snapshot. Do not ask, do not wait.
+- If the file exists, read it instead of rescanning the codebase.
 - Rescan and refresh this file only when the user explicitly asks.
 
 Expected snapshot structure (keep concise and high-level):

package/templates/.claude/rules/agentic-workflow.md

@@ -6,7 +6,7 @@ Follow these rules in every session.
 ## Read order at session start
 1. Read `AGENTS.md` — determines repo mode (single-repo vs meta-repo) and project context
 2. Read `.agent-workflow/README.md`
-3. Read `.agent-workflow/codebase.md` if it exists
+3. Read `.agent-workflow/codebase.md` — **if it does not exist, create it immediately by scanning the codebase before continuing**
 4. Read the target spec and any linked feature doc before planning or implementation
 5. Read relevant existing code (in the appropriate directory) before proposing structural changes
 
@@ -40,7 +40,7 @@ When asked to prepare a spec:
 - Default `acceptance_proof: true` unless user explicitly says otherwise
 - Collect supporting assets into `.agent-workflow/specs/assets/<spec-slug>/`
 - Prefer assets over embedding large JSON, CSV, or screenshots inline
-- When spec is ready, commit the changes, push the current branch `agent/spec/<slug>`, and open or update a GitLab MR (always, unless the user explicitly says to skip GitLab flow)
+- When spec is ready, commit the changes. If running non-interactively, also push the branch and open or update a GitLab MR automatically. If running interactively, stop after the commit and ask the user whether to push and open/update the MR before proceeding.
 
 ### B. Plan implementation
 When asked to implement a spec:

package/templates/.claude/skills/prepare-spec/SKILL.md

@@ -71,9 +71,16 @@ Use filenames that reveal purpose, for example:
 ## GitLab handoff behavior
 1. Stage only the relevant spec, asset, and doc changes.
 2. Create a focused commit.
-3. Push the current branch with `scripts/agent/git-push-branch.sh`.
-4. Generate MR body content using `scripts/agent/mr-template.sh spec <spec-path> - - -`.
-5. Open or update the MR using the scripts in `scripts/agent/`.
+3. Decide whether to proceed with push and MR:
+   - **Non-interactive session** (invoked with `-p`, no user to prompt) → proceed to steps 4–6 automatically.
+   - **Prompt includes `with push`** (e.g. "use prepare-spec skill with push …") → proceed to steps 4–6 automatically.
+   - **Prompt explicitly instructs full handoff** (e.g. "Full GitLab handoff — push the branch and open or update the MR without asking.") → proceed to steps 4–6 automatically.
+   - **All other interactive sessions** → stop here, report the commit, and ask:
+     > "Spec committed on `<branch>`. Should I push the branch and open/update the MR? (yes / no)"
+     Proceed to steps 4–6 only if the user confirms.
+4. Push the current branch with `scripts/agent/git-push-branch.sh`.
+5. Generate MR body content using `scripts/agent/mr-template.sh spec <spec-path> - - -`.
+6. Open or update the MR using the scripts in `scripts/agent/`.
 
 ## Final handoff
 Return:

package/templates/AGENTS.md

@@ -11,7 +11,7 @@ See `agents-workflow-dev-process.md` for both the **scripted workflow** (bash sc
 ## Quick Start
 
 1. Read `.agent-workflow/README.md` (workflow process — **Phase A** spec authorship vs **Phase B** delivery).
-2. Read `.agent-workflow/codebase.md` (codebase snapshot — do NOT rescan unless explicitly asked).
+2. Read `.agent-workflow/codebase.md` (codebase snapshot). **If it does not exist, create it immediately by scanning the codebase — do not skip this step.** Do NOT rescan if it already exists unless explicitly asked.
 3. Determine the task:
    - **Prepare a spec** (user pasted Notion URL, asked for chat-assisted drafting, or asked to convert Notion → spec): follow **Phase A**; write `.agent-workflow/specs/*.md` with `status: draft` unless the user says otherwise; use `.agent-workflow/notion-spec-mapping.md` for Notion.
    - **Implement a feature**: only specs with **`status: open`**; check `.agent-workflow/specs/` for those.

package/templates/scripts/agent/prepare-spec.sh

@@ -2,19 +2,21 @@
 set -euo pipefail
 
 INTERACTIVE=false
+PUSH=false
 
 # Parse flags
 POSITIONAL=()
 for arg in "$@"; do
   case "$arg" in
     --interactive) INTERACTIVE=true ;;
+    --push) PUSH=true ;;
     *) POSITIONAL+=("$arg") ;;
   esac
 done
 set -- "${POSITIONAL[@]}"
 
 if [[ $# -lt 2 ]]; then
-  echo "Usage: $0 [--interactive] <slug> <notion-url|gdrive-url|local-file>" >&2
+  echo "Usage: $0 [--interactive] [--push] <slug> <notion-url|gdrive-url|local-file>" >&2
   exit 1
 fi
 
@@ -39,8 +41,13 @@ echo "Worktree: ${WORKTREE_DIR}" >&2
 cd "${WORKTREE_DIR}"
 
 if [[ "$INTERACTIVE" == true ]]; then
-  echo "Starting interactive Claude session (GitLab flow disabled)..." >&2
-  exec claude --dangerously-skip-permissions "${SOURCE_PROMPT}. Skip GitLab flow — do not push or open an MR."
+  if [[ "$PUSH" == true ]]; then
+    echo "Starting interactive Claude session (full GitLab handoff enabled)..." >&2
+    exec claude --dangerously-skip-permissions "${SOURCE_PROMPT}. Full GitLab handoff — push the branch and open or update the MR without asking."
+  else
+    echo "Starting interactive Claude session..." >&2
+    exec claude --dangerously-skip-permissions "${SOURCE_PROMPT}"
+  fi
 else
   exec claude --dangerously-skip-permissions -p "${SOURCE_PROMPT}"
 fi