Skip to main content

Agent Overview

The SSSC Planner is a phase-based conversational agent that produces supply chain security assessments, standards mappings, gap analyses, and backlog handoff artifacts. It inventories capabilities across hve-core and physical-ai-toolchain, maps them to OpenSSF standards, and generates improvement projections.

Architecture

The agent follows six SSSC-specific instruction files, each scoped to a specific phase concern. It also references five shared instruction files from the Security Planning collection (identity, operational-buckets, standards-mapping, backlog-handoff, security-model) for cross-cutting patterns.

State Management

All state lives in .copilot-tracking/sssc-plans/{project-slug}/state.json. The agent follows a six-step protocol on every turn:

StepAction
READLoad the current state file
VALIDATEConfirm the state schema is intact and the current phase is consistent
DETERMINEDecide which phase and step to execute based on state and user input
EXECUTEPerform the phase work (questions, analysis, artifact generation)
UPDATEModify the in-memory state to reflect completed work
WRITEPersist the updated state back to the file

State Fields

The state file tracks 17 fields across scoping, analysis, and handoff concerns.

FieldTypeDescription
projectSlugstringKebab-case project identifier
ssscPlanFilestringPath to the main SSSC plan markdown file
currentPhasenumberCurrent phase (1-6)
entryModestringcapture, from-prd, from-brd, or from-security-plan
scopingCompletebooleanWhether Phase 1 scoping has been completed
assessmentCompletebooleanWhether Phase 2 capability inventory is complete
standardsMappedbooleanWhether Phase 3 standards mapping is complete
gapAnalysisCompletebooleanWhether Phase 4 gap analysis is complete
backlogGeneratedbooleanWhether Phase 5 backlog generation is complete
handoffGeneratedobject{ado: boolean, github: boolean}
context.techStackstring[]Target repository technology stack
context.packageManagersstring[]Package managers in use
context.ciPlatformstringCI/CD platform (GitHub Actions, Azure Pipelines, etc.)
context.releaseStrategystringRelease strategy (tags, branches, etc.)
context.complianceTargetsstring[]Compliance frameworks being targeted
referencesProcessedstring[]Paths to PRD/BRD/security-plan artifacts consumed
nextActionsstring[]Pending actions for the current or next phase
userPreferencesobjectAutonomy preference: full, partial, or manual
ssscEnabledbooleanWhether SSSC planning is active
securityPlannerLinkstringPath to the upstream Security Planner state file
raiPlannerLinkstringPath to an associated RAI Planner state file

Interaction Model

The agent follows strict question rules during each phase:

GuardrailDescription
3-5 questions per turnEnough to make progress without overwhelming the user
Emoji checklistsQuestions use ❓ for pending, ✅ for answered, and ❌ for blocked items
No phase advance without confirmationThe agent summarizes phase findings and asks for explicit approval before moving to the next phase

Session Resume

When a conversation resumes from a prior session, the agent follows a four-step recovery protocol:

  1. Read the state file from .copilot-tracking/sssc-plans/{project-slug}/.
  2. Validate that the state schema matches the expected version.
  3. Present a summary of completed phases and pending work.
  4. Continue from the current phase without re-asking answered questions.

A five-step post-summarization recovery handles cases where conversation context was compacted by the chat system.

Operational Constraints

  • All generated files are placed under .copilot-tracking/sssc-plans/{project-slug}/.
  • The agent never modifies source code or files outside its tracking directory.
  • The Researcher Subagent is dispatched for WAF/CAF runtime lookups when cloud-hosted components are in scope.
  • Cross-agent links (securityPlannerLink, raiPlannerLink) are populated but the agent does not force handoff to other agents.
File typeLocation
Agent.github/agents/security/sssc-planner.agent.md
Prompts.github/prompts/security/sssc-*.prompt.md
Instructions.github/instructions/security/
State.copilot-tracking/sssc-plans/{project-slug}/state.json

🤖 Crafted with precision by ✨Copilot following brilliant human instruction, then carefully refined by our team of discerning human reviewers.