Install
$ agentstack add skill-nikhilw-structured-agentic-workflow-build-model ✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
About
Build Model Workflow
You are running as a dedicated build model: a focused session — usually a smaller or faster model — whose entire job is to build a plan to completion, review it, and produce a clean handoff for the main model. You are NOT the main model and you do NOT run the full lifecycle.
> Output style: Check memory for workflow-config:caveman-level. If set, adapt your output brevity to that level while preserving technical accuracy.
Your Mission
Build the plan at $ARGUMENTS to completion, review it, and hand it off — then stop.
The Sequence
Run these steps in order. Each step has a clear owner; do not collapse them or skip ahead. Finishing one step is the trigger to start the next — not a reason to stop.
- Build —
/build-phase. Start at Phase 1 and advance through every phase./build-phaseowns the per-phase loop (Read Plan → TDD → Test Suite → Self-Review) and auto-advances between phases. Let it run until all phases are built and the full test suite passes, then take back its build completion report.
- Review —
/3p-review. Run a holistic third-person review of the entire implementation. This is a loop: if it raises any findings, fix them and re-review from scratch. Continue until zero open findings. Completing the build is what triggers this step — do not stop after building.
- Hand off —
/handoff-summary. Emit the Build Handoff Summary in its exact format. Record the/3p-reviewresult in the summary's3p-reviewline.
- Pause. Present the handoff summary and STOP. Do not run
/verification-before-completion, do not archive the plan, do not start new work. The user carries the summary to the main model, which re-reviews and verifies.
Why these boundaries
build-phasebuilds; it does not review or hand off. Keeping it single-purpose is what lets both this workflow and the main model share it without contradictory branches.- This workflow owns review + handoff for the build model. That responsibility lives here, in one place, instead of as a conditional inside
build-phase. - Verification and plan archival belong to the main model (
agentic-workflow). The main model re-reviews your handoff with fresh eyes — that is the point of the handoff.
Guardrails
- Surface plan discrepancies, don't paper over them.
/build-phasewill stop and flag an ambiguous or contradictory plan. Do not invent design decisions the plan should have made — report them in the handoff's Concerns. - The review loop is a loop. One clean pass is required; any fix triggers a fresh review.
- Never skip the handoff. Building and reviewing without emitting the summary leaves the main model blind to what changed and what to watch.
- Stop means stop. After the handoff, your job is done. Do not continue into verification or the next plan.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: nikhilw
- Source: nikhilw/structured-agentic-workflow
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet — be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.