plan — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited plan (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
This skill is the explicit gate version of stratified planning: you invoke it deliberately when you want a plan as a reviewable artifact before any code changes. The same shape is also produced ambiently by the trabuco-planner subagent when CLAUDE.md's delegation rule fires — this skill is for cases where you want to drive the gate yourself.
A markdown plan in canonical Trabuco-stratified shape:
# Plan: <task>
## Stage 1 — Model
- Files added/modified
**Validate:** `mvn -pl Model install`
## Stage 2 — Jobs (or whichever Layer-1 contract module applies)
...
## Stage 3 — SQLDatastore (or NoSQL)
...
(continues through layers present in this project)
## Final verification
- mvn install across the reactor
- trabuco doctor reports cleanFor non-trivial features (3+ stages), the plan is also saved at .trabuco/plans/<date>-<slug>.md so the PR description can reference it and reviewers can compare diff to plan.
mcp__trabuco__get_project_info(or cat .trabuco.json | jq .modules if MCP isn't available). If the user is not in a Trabuco project, instruct them to either cd in or run trabuco init first — stratified planning doesn't apply outside Trabuco's hierarchy.
plan-production logic: archetype detection (additive feature / job feature / event feature / rename / bug fix / single-module / migration-only), per-stage validate commands, depends-on rationale, anti-pattern avoidance.
Invoke as: ask the main agent to spawn @trabuco-planner with the user's task description. The planner reads .trabuco.json, picks the archetype, and produces the plan.
📋 Plan ready. Approve to execute, request changes, or cancel?agent works through the stages in order, running each Validate: command before proceeding to the next stage.
what's wrong; refine the plan; re-render for approval. Don't execute until the plan is right.
before code lands
plan-first workflow
archetype, where the contract/consumer boundary falls) before any code is written
and the ceremony is wasteful
phase, not for retroactive documentation
point, not the implementation. The planner persona is where the archetype-detection and convention-checking logic lives.
approval, only then hand off to the main agent for code changes.
.trabuco/plans/ so the PR has aversioned record of the design intent.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.