overview glossary

Glossary

Template repo - This repository (intel-agency/agent-context). The substrate from which downstream instances are cloned.
Clone instance - A repository created from this template, seeded with a unique application plan. Any repo other than agent-context is a clone instance.
Downstream clone - Same as clone instance. Used when describing the cloning pipeline.

Plan - The top-level issue in the hierarchy. One per project. Title format: Plan: <Name>.
Epic - A major work group within a plan. Often maps to a "Phase" in the source plan document. Title format: Epic <N>: <Name>.
Story - An atomic, independently-completable work item within an epic. Title format: Story <N>.<M>: <Name>.
Task - An optional sub-tier beneath stories, only when the plan nests a sub-tier. Title format: Task <N>.<M>.<K>: <Name>.
Sub-issue - GitHub's mechanism for parent/child issue linking. The hierarchy uses sub-issues exclusively, never issue-linked task lists.

Memory - Durable project context stored at .agents/memory.md. Four sections: Current Activity, Completed Work Items, Decisions, Remember To Do.
Rules - Subject-specific convention files under .agents/rules/. One file per topic (tools, validation, source control, etc.).
Skill - A self-contained agent capability following the Agent Skills specification. Lives under .agents/skills/ with SKILL.md, scripts/, references/, and assets/.

Build step - Markdown linting and relative link validation.
Scan step - PSScriptAnalyzer static analysis (error severity) and gitleaks secret scanning.
Test step - Pester test suite with code coverage and an 85% threshold gate.
JaCoCo - Coverage output format (XML) produced by Pester's code coverage feature.
ReportGenerator - Tool that converts JaCoCo XML into browsable HTML coverage reports.

OpenCode - The agent runtime that reads opencode.jsonc for model providers, MCP servers, and agent definitions.
Agent definitions - YAML-like markdown files under .opencode/agents/ defining agent roles (orchestrator, team-lead, developer, etc.).
MCP servers - Model Context Protocol servers configured in opencode.jsonc for tools like sequential-thinking, memory-graph, web search, and code reading.
Workflow assignments - ShortId-to-URL mappings in local_ai_instruction_modules/ai-workflow-assignments.md that resolve /commands to canonical instruction files.
Dynamic workflows - ShortId-to-URL mappings in local_ai_instruction_modules/ai-dynamic-workflows.md for parameterized workflow templates.

Forensic log - Per-run log file (gh-init-<slug>-<timestamp>.log) recording every GitHub API call and operation outcome for post-execution analysis.
DryRun - Preview mode for all operation scripts. Shows what would be created without making any API calls.
Filler detection - Automated check that flags body sections byte-identical across 3+ sibling issues, indicating template boilerplate was pasted instead of plan-derived content.
Orchestration signal label - gh-issue-tracking:init-success label applied to the Plan issue on a successful apply run. A downstream orchestrator matches it to know the hierarchy is ready.

App stack - A pre-defined language and tech stack profile in .agents/rules/app-stacks/, referenced from app development plans to specify the language, tools, and packages to use. Named by slug ID (e.g. dotnet-aspire-aspnet-blazor).