Understanding Workflow Prompts

Agentic capabilities in SD Elements are delivered through Workflow Prompts. While the MCP server provides the underlying tools, a Workflow Prompt defines the higher-level orchestration—how those tools are used together to accomplish a defined outcome. A Workflow Prompt guides an agent through structured, multi-step, repeatable reasoning across SD Elements concepts, replacing fragmented manual steps or brittle scripts with governed automation.

What Are SD Elements AppSec Skills for Agentic AI Workflow?

Skills are an open framework for packaging AI agent instructions and tools into portable, reusable playbooks. Rather than loading all instructions upfront, a skill allows an agent to dynamically pull in countermeasure-specific logic only when it’s needed — keeping the agent focused and its context lean.

SD Elements Skills connect AI agents to the SD Elements library of security and compliance requirements, enabling them to enforce your organization’s controls automatically during software development. Any AI agent that supports the open Model Context Protocol (MCP) can integrate your security policies directly into its development workflows.

Three types of artifacts work together to make this possible:

SKILL.md — Countermeasure Skills

A SKILL.md file is an open standard for defining a reusable skill for an AI agent. It contains the instructions, metadata, and resources that teach an agent how to perform a specific, well-scoped task — such as "get latest countermeasures" or "apply a security fix."

In SD Elements, Countermeasure Skills are task-level skills that specify requirements and implementation guidance for a particular security control or compliance requirement, drawn from the SD Elements library. These are loaded into the agent’s context temporarily, only as needed, keeping overhead minimal.

Generated SKILL.md files are stamped with a status at creation time. A status of Pending is expected and does not indicate an error — the terminal status (for example, Done) is set later, once /apply-security-fixes has run against that countermeasure.

AGENTS.md — Weakness Remediation Plan

An AGENTS.md file is an open standard for static, project-level agent instructions: code conventions, commands, and rules that ensure any agent working on a repository behaves consistently.

In SD Elements, workflow prompts generate a project-specific Weakness Remediation Plan as an AGENTS.md file — an enforceable contract that indexes the relevant countermeasure skills dictated by your team’s risk policy into a plan a coding agent can execute. Because AGENTS.md lives in the repository, it becomes part of the persistent project context: the agent is always aware of it. Teams can also edit and adapt the plan directly in their repository to suit their needs.

Key difference: SKILL.md content is loaded on demand, only when the agent needs it. AGENTS.md is always present in the project context, making it the persistent contract the agent operates under.

Workflow Prompts — Top-Level Orchestration

Workflow Prompts are static, top-level commands available directly via the SD Elements MCP server. They define the higher-level orchestration — how tools, skills, and agent interactions are combined to accomplish a complete outcome. A workflow can trigger multi-step processes, call other tools, and interact with the user via chat or the command line to clarify tasks along the way.

How They Work Together

Artifact Type Scope Persistence

Workflow Prompt

Top-level command

Full workflow orchestration

Static, served via MCP

SKILL.md

Countermeasure Skill

Single task or control

Loaded on demand

AGENTS.md

Weakness Remediation Plan

Entire project

Always present in repo

The flow in practice:

  1. A Workflow Prompt drives the overall process and determines which countermeasures apply.

  2. Countermeasure Skills (SKILL.md) are pulled in on demand to provide specific security guidance as each control is addressed.

  3. The Weakness Remediation Plan (AGENTS.md) is written to the repository as a persistent contract, keeping the agent aligned with your project’s risk policy throughout the project lifecycle.

The Agentic Workflow Pipeline

The full workflow is composed of four stages, each a separate Workflow Prompt. Each stage hands off its state to the next automatically via an internal handoff file — you do not need to pass any data between prompts manually.

Stage Workflow Prompt What it does

1. Survey

/setup-security-plan-from-repo
/create-plan-from-specs

Analyzes your repository or specs, completes the SD Elements survey, and commits it.

2. Generate

/generate-security-skill-files

Fetches applicable countermeasures, classifies them, and generates AGENTS.md plus per-countermeasure SKILL.md files.

3. Apply

/apply-security-fixes

Implements the fixes described in the generated skill files.

4. Verify

/code-scan-verification-validation
/zp-code-scan-verification-validation

Confirms fixes are in place in the codebase and updates countermeasure status in SD Elements.

The Survey stage (/setup-security-plan-from-repo and /create-plan-from-specs) ends once the survey is committed. It does not fetch countermeasures or generate skill files — that is handled by /generate-security-skill-files, a separate prompt. If you are only interested in survey results, running the Survey stage alone is sufficient; you do not need to run the later stages.

You can stop after any stage depending on what you need. Each prompt is designed to be run in its own session (context window) — see Usage Notes.

Available Workflows

Setup Security Plan from Existing Codebase - /setup-security-plan-from-repo

Analyze a source code repository to create an SDE project, collect applicable security requirements in SD Elements, and commit a completed survey as the foundation for the agentic security mitigation plan.

Key Functions:

  • Checks whether the current workspace is already the target repository before offering to browse elsewhere.

  • Asks whether this is an initial assessment or an update to an existing project, and supports an optional version label for update runs.

  • Completes SD Elements survey with evidence from repository analysis.

  • Commits the survey and hands off to /generate-security-skill-files for countermeasure retrieval and skill-file generation.

  • Assigns risk policy after the survey is committed (see Risk Policy Assignment).

  • Preserves existing project content, avoiding overwriting prior configuration.

This workflow prompt is typically the starting point for teams working from an existing codebase. It ends at survey commit — run /generate-security-skill-files next to retrieve countermeasures and generate skill files.

Create a Security Plan from Design Documents - /create-security-plan-from-specs

Interpret application specifications to complete an SD Elements survey and lay the groundwork for an agentic risk-mitigation plan before full implementation.

Key Functions:

  • Processes SDLC documents (PRDs, requirements, design, architecture).

  • Supports sources: local files, Confluence, or Jira.

  • Completes SD Elements project survey using specification content.

  • Commits the survey and hands off to /generate-security-skill-files for countermeasure retrieval and skill-file generation.

  • Assigns risk policy after the survey is committed (see Risk Policy Assignment).

This workflow prompt is designed for greenfield projects and early design-stage workflows, allowing teams to begin threat modeling before code exists. Like the repo-based workflow, it ends at survey commit — run /generate-security-skill-files next.

Generate Security Skill Files - /generate-security-skill-files

Once a survey has been committed by either survey workflow, this prompt retrieves the project’s applicable countermeasures, classifies them, and generates the artifacts an agent needs to act on them.

Key Functions:

  • Fetches the project’s countermeasures following survey commit.

  • Lets you select scope — all countermeasures, specific ones, or a free-text search.

  • Classifies each countermeasure as CODE_FIX, INFRA, or PROCESS (see Fix Classification Logic)

  • Generates the root AGENTS.md index and per-countermeasure skills/{domain}/{countermeasure-slug}/SKILL.md files.

  • Creates the security hardening branch and a local backup of prior AI configuration immediately before writing generated files.

  • Hands off to /apply-security-fixes once file generation is complete.

This workflow prompt is used after either survey workflow and before /apply-security-fixes.

Apply Security Fixes to Codebase - /apply-security-fixes

Leveraging AI SAST, the system reviews changes within the repository to suggest iterative updates. This ensures that project requirements remain consistently aligned with the evolving codebase.

Key Functions:

  • Reads the generated skill files and applies fixes to source code.

  • Supports flexible fix application: broadly or in controlled steps.

  • Allows incremental work with individual change review/approval.

  • Records progress to SD Elements with notes for visibility and auditability.

  • Supports resuming interrupted work.

  • Provides clearer progress reporting.

How Fixes Are Classified

Countermeasure classification happens in /generate-security-skill-files, before any files are generated. Each countermeasure is classified as CODE_FIX, INFRA, or PROCESS:

Category Meaning What gets generated

CODE_FIX

Requires a change to application code within the repository.

A per-countermeasure SKILL.md file with implementation guidance, tracked in AGENTS.md.

INFRA

Requires an infrastructure or configuration change outside the repository (e.g. server hardening, TLS configuration).

A doc-only SKILL.md file, since the change happens outside the codebase; tracked in AGENTS.md.

PROCESS

An organizational item with no code or infrastructure component (e.g. review policies, breach notification procedures).

No file is generated. The item is recorded as a note directly in SD Elements.

If a countermeasure should be marked Not Applicable but wasn’t, this usually means the agent lacked sufficient context to make that determination. Adding more detail to your survey answers, or to repository documentation the agent can read, generally improves classification accuracy.

This workflow prompt is best used after /generate-security-skill-files has produced skill files to work from.

Verify Security Countermeasures via AI Code Scan - /code-scan-verification-validation

Using AI-powered static analysis using your model of choice, the system verifies that each countermeasure is actually mitigated in the source code and records structured Verification Notes to SD Elements for auditability and compliance evidence.

Key Functions:

  • Reads countermeasure context from SD Elements (vulnerability, fix guidance, how-tos).

  • Performs AI code analysis against relevant source files for each countermeasure.

  • Derives a verdict (pass, partial, or fail) with confidence level and cited findings.

  • Post Verification Notes to the task Evidence section in SD Elements.

  • Supports two modes: standalone (baseline scan before fixes) and handoff (post-fix confirmation).

  • Supports flexible scope: verify all, verify specific countermeasures, or verify one-by-one interactively.

  • Countermeasures that have been verified as passed are optionally marked as DONE in SD Elements.

This workflow prompt is used twice in the recommended flow—post-planning for a baseline and post-fix for confirmation—this prompt provides objective, auditable evidence of mitigation status.

Verify Countermeasures with a Security Scanner - /verify-with-scanner

Verify an SD Elements project’s security countermeasures by running an out-of-the-box security scanner (e.g. SonarQube) through SD Elements' analysis integration. SD Elements runs the scan import, maps findings to countermeasures, and writes Verification Notes server-side — you never call the scanner directly and never handle scanner credentials.

Before running this prompt, an administrator must configure the scanner as a global Verification connection in SD Elements, under System > Integration. That global connection holds the server address and credentials; the prompt only creates the project-level connection that inherits them, which is why the agent never needs a scanner token. See Verification integrations for setup.

This prompt and the verification_integration tool are only available when the integrations feature is enabled on the MCP server. Set a features: integrations request header for a hosted MCP server, or MCP_FEATURES=integrations in the environment for a local one. Without it, the prompt is not listed and requesting it by name returns "Prompt not found".

Key Functions:

  • Resolves or creates a project connection to a live-connector scanner, reusing an existing connection rather than creating a duplicate.

  • Distinguishes remote (live) connections from file-upload connections, which this workflow cannot trigger.

  • Triggers a scan import and polls the sync job until it completes.

  • Reads back Verification Notes and reports a pass, partial, or fail outcome per countermeasure, with finding evidence and counts.

  • Surfaces scanner findings that didn’t map to any countermeasure in the project, so nothing is silently dropped.

  • Optionally pins a verification result as the authoritative outcome for a countermeasure.

This workflow prompt is the live-connector counterpart to AI-based code scan verification, connecting and scanning with out-of-box plugins that you have set up within SD Elements.

This workflow drives remote connections only — those that pull from a live scanner. It cannot run a file-upload connection, where a person attaches a scan report by hand, because there is no live endpoint to trigger. If a project’s only verification connection is a file-upload one, the agent says so instead of attempting a scan that would fail, and points you to the path that does work: the project’s Integration > Verification tab, then CONNECTION, then pick the scanner under File Uploads, attach the report, and click CREATE.

Generate Issue Tracker Tasks from an SD Elements Project - /generate-issue-tracker-tasks

Sync an SD Elements project’s security countermeasures into an issue tracker (e.g. Jira) by driving SD Elements' own out-of-the-box issue-tracker integration. SD Elements performs the sync, deduplication, and write-back server-side — you never call the tracker directly and never handle tracker credentials.

Before running this prompt, an administrator must configure the tracker as a global Issue Tracker connection in SD Elements, under System > Integration. That global connection holds the server address and credentials; the prompt only creates the project-level connection that inherits them, and asks you for the tracker’s project or board key. See Issue Tracker Integration for setup.

This prompt and the issue_tracker_integration tool are only available when the integrations feature is enabled on the MCP server. Set a features: integrations request header for a hosted MCP server, or MCP_FEATURES=integrations in the environment for a local one. Without it, the prompt is not listed and requesting it by name returns "Prompt not found".

Key Functions:

  • Resolves or creates a project connection to a global issue-tracker connector, reusing an existing connection when present.

  • Tests the connection before syncing and surfaces configuration errors verbatim so they can be corrected.

  • Triggers a sync and polls the job until it completes, waiting rather than starting a duplicate if one is already running.

  • Reports created, updated, already-in-sync, and failed counts, with tracker links for created or updated tasks.

  • Accounts for every in-scope countermeasure, explicitly listing any that couldn’t be mapped or synced.

This workflow prompt is best used once security countermeasures are established in a project, to keep issue tracker tasks aligned with the SD Elements plan on an ongoing basis.

Risk Policy Assignment

Both survey workflow prompts (/setup-security-plan-from-repo and /create-plan-from-specs) create the project without a risk policy attached. Risk policy is assigned in a separate step after the survey is committed:

  • If a risk policy is already set on the project, the workflow displays it and asks whether to keep it.

  • If none is set, you are asked to select one before /generate-security-skill-files fetches countermeasures.

  • If no policy is selected at all, SD Elements automatically applies its default policy for the business unit or organization.

Restricting this step: If your organization requires that certain teams (for example, Product Teams) not be able to change risk policy, you can customize the relevant SKILL.md file to skip the selection step and always inherit the existing or default policy instead of prompting. See Customize Skills for guidance on modifying prompt behavior.

Customize Skills

Workflow Prompts will be managed in the existing SD Elements content library, which makes them easy to customize, manage, and distribute across teams. Because these particles skills are defined in simple Markdown (SKILL.md) files, they are fully transparent and customizable.

  • Edit Existing Skills: You can adjust the instructions, add new verification steps, or modify the MCP tool calls in any existing skill to better match your team’s specific processes.

  • Create Custom Workflows: You can easily create your own SKILL.md files from scratch to automate other security tasks or create entirely new workflows tailored to your organization’s needs. This allows you to build a library of custom, automated security processes that your team can rely on.

Will my customizations be overwritten?

Platform updates to workflow prompts and skills are documented on the What’s new in the agentic AI workflow page with every release. If a release changes a prompt or skill you have customized, it will be called out there. We recommend checking that page before a new release reaches your environment, and comparing your local modifications against the release notes if you’re unsure whether your customization is still current.

For the best results:

  • Each workflow prompt should be used in the agents' plan mode for it to prepare an execution plan to run the workflow prompt.

  • Each workflow prompt is best used in a separate session (context window).

Usage Notes

  • Best used for accelerating project planning and ongoing alignment.

  • Outputs will vary depending on the model used and the agent’s capabilities.

  • Frontier models typically perform better and provide the best results.

  • To build your own Skills, refer to this documentation for assistance.

results matching ""

    No results matching ""