| name | orchestrator-inject |
| description | Commit and inject a chunk into the orchestrator for background execution. |
Tips
- The ve command is an installed CLI tool, not a file in the repository. Do not
search for it - run it directly via Bash.
Instructions
Inject a chunk into the orchestrator for background execution, with a pre-flight
commit check to ensure the chunk files are available in the worktree.
Step 1: Argument Parsing
Input: $ARGUMENTS
Parse the arguments for an optional chunk name (positional).
Example invocations:
/orchestrator-inject my_chunk
/orchestrator-inject
If no chunk name is provided, run ve chunk list --current to resolve the
current IMPLEMENTING chunk. If that returns nothing, run ve chunk list and
look for FUTURE chunks. If still ambiguous, ask the operator which chunk to
inject.
Store the resolved chunk name as <chunk> for subsequent steps.
Step 2: Pre-flight — Ensure chunk is committed
Run git status --porcelain docs/chunks/<chunk>/ to check if GOAL.md or
PLAN.md have uncommitted changes (modified, untracked, etc.).
If changes exist:
-
Stage the chunk files:
git add docs/chunks/<chunk>/GOAL.md docs/chunks/<chunk>/PLAN.md
-
Commit with a conventional message:
git commit -m "docs: commit <chunk> for orchestrator injection"
-
Report to the operator that files were auto-committed.
If the chunk files are already committed and clean:
Report that the chunk files are already committed and skip to Step 3.
If git is not available (the git status command fails):
Skip the commit step and proceed to Step 3. The orchestrator requires git
for worktrees, so this scenario is unlikely but should be handled gracefully.
Step 3: Ensure orchestrator is running
Run ve orch status. If the orchestrator is not running, start it with
ve orch start.
Step 4: Inject the chunk
Run ve orch inject <chunk> and capture the output. Report success or
failure to the operator.
Step 5: Offer monitoring
After successful injection, suggest:
Would you like me to monitor this chunk? I can run
/orchestrator-monitor <chunk>.
Do not auto-start monitoring unless the operator confirms.