workflow-blueprint
Blueprint workflow to scaffold agent-oriented documentation with progressive disclosure and executable contracts.
How do I install this agent skill?
npx skills add https://github.com/devton/agentic-workflow-blueprint --skill workflow-blueprintIs this agent skill safe to install?
- Gen Agent Trust Hubpass
This skill is a comprehensive meta-blueprint designed to scaffold agentic workflows and documentation. It provides structured templates for security auditing, CI/CD hardening, and infrastructure operations. It is generally safe but carries a low-risk surface for indirect prompt injection as it is designed to process external inputs such as git diffs and system logs.
- Socketwarn
1 alert: gptSecurity
- Snykpass
Risk: LOW · No issues
What does this agent skill do?
agentic-workflows.blueprint
Goal
Provide a reusable blueprint workflow that scaffolds an "agentic workflows" documentation structure for any codebase, with progressive disclosure and executable contracts (workflows).
Scope
- Applies to: any repository that wants an agent-oriented documentation system (orchestrator + workflow skills + references + runbooks).
- Does not cover: implementing product features; this is scaffolding/documentation only.
Triggers
- "Create agentic workflow structure"
- "Set up skills folder + workflows"
- "Make docs agent-friendly"
- "Refactor AGENTS.md into linked skills"
Inputs
projectSlug: short identifier for the repo (e.g.my-backend)baseBranch: default integration branch (e.g.develop,main)techStack: short list (e.g.NestJS + MikroORM + Graphile Worker)existingRootDoc: root instruction file path (AGENTS.md,CLAUDE.md, etc.)workflowsWanted: list of workflow ids to scaffold (e.g.modules,specs,document,network-engineering,infra-operations,iac,os-platform,plan-to-blueprint)constraints: project hard rules (e.g. "no emojis", "mock external boundaries only")skillName(optional): slash-invocation name for the project entry skill; defaults toprojectSlug(lowercase, hyphens only)
Outputs
Creates a minimal, navigable structure:
skills/<projectSlug>/
SKILL.md
template.json ← Declarative manifest (commands, entry)
reference/
routing-matrix.md
role-contracts.md
hook-blueprint.md (optional)
workflows/
<workflowName>/
SKILL.md
template.json (optional) ← Per-workflow manifest when useful
docs/runbooks/
agent-role-system.md
agent-role-hooks.md (optional)
plan-to-blueprint.md (when workflow included)
And updates the root doc (AGENTS.md or equivalent) to link to the new entrypoints.
This blueprint folder carries bundled workflows (embed-aihero-radioactive, mattpocock/wayfinder, mattpocock/grilling, mattpocock/domain-modeling, mattpocock/research, mattpocock/prototype, mattpocock/to-spec, radioactive, brainstorming, plan-writing, ui-ux-pro-max, remotion-video-motion, thermo-nuclear-code-quality-review, thermo-fix, changelog-generator, document, review, changelog, linear, mcp-linear-planner, mcp-linear-sync, network-engineering, infra-operations, iac, os-platform, implementing-devsecops-security-scanning, scanning-containers-with-trivy-in-cicd, scanning-docker-images-with-trivy, scanning-kubernetes-manifests-with-kubesec, implementing-network-policies-for-kubernetes, implementing-rbac-hardening-for-kubernetes, implementing-pod-security-admission-controller, securing-aws-iam-permissions, securing-container-registry-images, securing-kubernetes-on-cloud, triaging-vulnerabilities-with-ssvc-framework, performing-kubernetes-cis-benchmark-with-kube-bench, analyzing-kubernetes-audit-logs, securing-github-actions-workflows, performing-container-image-hardening, remediating-s3-bucket-misconfiguration, performing-container-security-scanning-with-trivy, performing-vulnerability-scanning-with-nessus, implementing-syslog-centralization-with-rsyslog, c4-architecture, plan-to-blueprint) to demonstrate full-lifecycle chained execution, Matt Pocock / AI Hero methodology embedding, plan-to-skill transformation, visual documentation generation, MCP integration patterns, and infrastructure operations coverage.
It can also carry runbook examples under runbooks/ to show operator-facing execution playbooks for those workflows.
Invariants (guardrails)
- Progressive disclosure: root doc stays short; details live behind links.
- Executable contracts: every workflow is written as a contract the agent can follow:
Goal,Scope,Triggers,Inputs,Invariants,Procedure,Outputs,Review gate,References.
- No duplication: do not copy/paste long rules across files; link to the source of truth.
- Consistency: workflow ids and file paths must match exactly across all references.
- Minimal surface: only add the workflows actually requested.
- Source of truth:
SKILL.mdis authoritative;template.jsonis a complementary declarative layer for agents/tools that expose command interfaces. - Command routing: router-only — users invoke
/<skillName> <cmd>; the entry skill resolvescmdtoworkflows/<cmd>/SKILL.mdand loads only that contract. - Manifest sync: every command in
template.jsonmust map to an existing workflow folder and matchrouting-matrix.md.
Procedure
1) Create the project entry skill
Create skills/<projectSlug>/SKILL.md as the global entrypoint:
- A short description of what the skill is for.
- An "orchestrator" section that explains:
- how to classify a task
- how to select one workflow
- how to close (validation gate if applicable)
- A Command routing section (router-only):
- invocation pattern:
/<skillName> <cmd>(e.g./my-backend document) - map each
cmdinworkflowsWantedtoworkflows/<cmd>/SKILL.md - if the agent does not parse subcommands, read
cmdfrom the user message and load the mapped workflow contract - do not create flat alias skills unless explicitly requested
- invocation pattern:
- A list of internal workflow helpers:
workflows/<name>/SKILL.mdlinks
- Project constraints and hard rules (short bullets).
Create skills/<projectSlug>/template.json as the declarative manifest:
{
"name": "<skillName>",
"version": "1.0.0",
"entry": "SKILL.md",
"routing": "router-only",
"commands": [
{
"name": "<workflowName>",
"description": "Short trigger description",
"skill": "workflows/<workflowName>/SKILL.md"
}
]
}
- Include one
commands[]entry per item inworkflowsWanted. - Keep
namealigned withskillName(folder may remainprojectSlug).
2) Create references (progressive disclosure)
In skills/<projectSlug>/reference/, create:
routing-matrix.md: task category -> workflow mapping; include slash form/<skillName> <cmd>alongside intent triggers.role-contracts.md: roles, boundaries, handoffs.- Optional
hook-blueprint.md: opt-in automation checklist.
Keep each file self-contained and linkable.
3) Scaffold each workflow as an executable contract
For each workflow in workflowsWanted, create:
skills/<projectSlug>/workflows/<workflowName>/SKILL.md with:
-
id:"<projectSlug>.workflow.<workflowName>"(as the first header line) -
Goal: 1 sentence -
Scope: applies/does not cover -
Triggers: file triggers + intent triggers -
Inputs: baseBranch, diff scope, required config -
Invariants: the project's hard rules + workflow-specific rules -
Procedure: deterministic steps (evidence-driven; use git diff when documenting) -
Outputs: what files/notes/checkpoints must be produced -
Review gate: checklist with pass/fail criteria -
References: links back toskills/<projectSlug>/SKILL.mdand any deep dives -
Special Scaffold Case —
embed-aihero-radioactive: When requested, adaptskills/<projectSlug>/workflows/radioactive/SKILL.mdto run the 14-phase lifecycle embedding Matt Pocock's skills, and scaffoldskills/<projectSlug>/workflows/mattpocock/containingwayfinder,grilling,domain-modeling,research,prototype, andto-spec. Ensure Phase 1 ALWAYS executes the primary project skill (skills/<projectSlug>/SKILL.md) FIRST for request classification and constraint extraction before executing/wayfinder.
Optionally, for each workflow, add workflows/<workflowName>/template.json when an external tool needs a standalone command descriptor; keep it minimal (name, entry, parent).
4) Wire everything into the root doc
Update existingRootDoc to include:
- "Start here": link to
skills/<projectSlug>/SKILL.md - Under "Skills" (or similar), list:
- the project skill link
- internal workflow skills links
- runbook links
Do not duplicate workflow contents in the root doc.
5) Optional: deprecate legacy skill locations (wrapper)
If there are existing skills in other directories:
- Keep the file
- Add a top banner:
- "Moved: canonical workflow is at
skills/<projectSlug>/workflows/..."
- "Moved: canonical workflow is at
- Leave the rest as a deep dive reference
6) Consistency verification (required)
Before declaring the scaffold done:
- Verify every link path exists (
.md). - Verify workflow ids are consistent:
projectSlug.workflow.*matches the file it lives in.
- Verify root doc points only to canonical locations.
- Verify
template.jsoncommands[].namevalues exist underworkflows/and appear in Command routing +routing-matrix.md. - Verify router-only behavior is documented in the project entry
SKILL.md(no undocumented flat aliases).
Review gate (must pass)
- Root doc remains minimal and only links out.
- Each workflow has the full contract sections (Goal..References).
- Constraints are explicit and testable (no vague "best practices").
- No duplication between root, project skill, and workflows.
- All links resolve (
.md). template.jsonis valid JSON and consistent withworkflowsWanted.- Command routing resolves every listed subcommand to exactly one workflow contract.
Notes
- This blueprint is intentionally stack-agnostic. For stack-specific rules (logging, ORM patterns, testing rules, network policies, IaC standards, and OS baselines), keep them in the project skill and link them from workflows.
- Installable skill contract with frontmatter: see
SKILL.md(name: workflow-blueprint).
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/devton/agentic-workflow-blueprint/workflow-blueprint">View workflow-blueprint on skillZs</a>