Guides
Guides45 pages

Guides

Task-focused guidance for setup, routing, prompt authoring, skills, security, validation, cleanup, and documentation operations.

Map

Choose by responsibility

InstallProject payload and runtime boundary
RouteKeyword scoring and scenario selection
AuthorPrompt bodies, scenarios, pairs, and skills
SecureAuthorization, evidence, and supply chain
VerifyTests, manifest, and CI gates
GUIDEAgent Library Integration GuideThis guide explains how MD capabilities connect to a separate agent library without duplicating the same ontology, prompt bodies, or governance rules in two places.→ GUIDEArchitecture GuideThis guide explains the suite as a set of cooperating layers. The architecture is designed to prevent a single oversized prompt from owning context, authority, research, execution,→ GUIDEAuto-Orchestration Runtime GuideThe automatic layer is a conditional compiler. It never loads all auto-prompts by default. It evaluates route-changing intent ambiguity, genuine skill need, missing capability, acq→ GUIDEAuto-Prompts and Conditional Routing GuideThe auto-prompt layer resolves a small set of recurring orchestration problems that should not be reimplemented inside every domain prompt:→ GUIDEBounded Loop Orchestration Guide`MD-197` can loop **any prompt, scenario, or exact skill**. The target is generic. The eligibility test—not the domain—decides whether looping is useful.→ GUIDECI and Testing GuideThis guide describes the deterministic checks that keep the suite coherent after prompt, schema, scenario, policy, skill, model, documentation, or runtime changes.→ GUIDEIdentity and Runtime Path GuideMission Directives exposes current, permanent capability identities and current agent skill destinations. The repository does not ship historical release maps, redirect tables, or → GUIDECompletion Criteria Guide`<completion_criteria>` is the contract that prevents a model or operator from confusing activity with completion. It answers:→ GUIDEContributor GuideThis guide explains how to change the suite without creating duplicate capabilities, weakening safety, breaking permanent identities, or allowing generated artifacts to drift.→ GUIDECross-Platform Tooling Guide[Back to Mastery Manual](MD_MASTERY_MANUAL.md) · [Manual index](MANUALS.md)→ GUIDEDocumentation Site GuideThe static Astro site publishes the public documentation surface at `https://manojpisini.github.io/mission-directives/`. The source lives under `site/` and uses the checked-in HTML→ GUIDEExecutive Decision Rules GuideExecutive prompts take approved action. Procedure alone is insufficient because execution encounters conflicts, incomplete budgets, stale evidence, failed checks, and domain-specif→ GUIDEGeneric Skill Execution Guide`MD-196` is a universal typed adapter for **any exact installed skill**. It is not limited to design skills, personal skills, or skills already curated in `skill_registry.json`. Th→ GUIDEGitHub Actions Failure History and Pre-Push GuideThis guide records every failed GitHub Actions validation run through July 28, 2026, its root cause, durable correction, and prevention rule. GitHub Actions results are commit-spec→ GUIDEgstack Integration MatrixThe suite treats every gstack skill as an optional adapter. Its loaded schema is authoritative; the prompt system remains responsible for evidence, permissions, output contracts, a→ GUIDEMission Directives Installation and Project Integration GuideMission Directives 2.0 installs a global command and a pinned, project-specific runtime. The global package provides lifecycle commands. Routing inside an initialized project uses → GUIDEInstalled Runtime Payload GuideThis guide defines the boundary between a working-project installation and the Mission Directives source repository. `config/runtime_payload.json` is the machine-readable source of→ GUIDEInstalled Skills Inventory Guide`installed_skills_inventory.json` records a snapshot of skills detected in the user's OpenCode global skill directory. The supplied snapshot was normalized from the OpenCode global→ GUIDELocal Output Viewer GuideThe local viewer is a visual wrapper over existing generated files. It does not rename, rewrite, index, or duplicate outputs. The installed templates and CSS live in `.mission-dire→ GUIDELocal Skill Registration GuideThis guide governs skills created locally rather than installed from a reviewed remote repository. Local creation is a fallback for a genuine capability gap, not a shortcut around → GUIDELogging and Telemetry Guide[Back to Mastery Manual](MD_MASTERY_MANUAL.md) · [Manual index](MANUALS.md)→ GUIDEMigration GuideMigration moves only content proven to belong to an older managed Mission Directives installation. Generic project folders are not treated as owned merely because their names match→ GUIDEModel Routing GuideModel routing selects a model only when measured evidence supports the requested prompt, evidence lane, output contract, tools, and assurance profile.→ GUIDEPackaging and Release Guide`pyproject.toml` defines the public `mission-directives` package, Python 3.11 minimum, `mission-directives` console entry point, runtime dependencies, and `MIT OR Apache-2.0` licen→ GUIDEPair Authoring and Verification GuideA genuine pair separates non-mutating investigation from authorized execution. The split is justified only when the handoff improves safety, reviewability, parallelism, or output q→ GUIDEPlan Review and Exact-Twin Execution GuideThis guide defines the mandatory transition from a paired investigative planning prompt to its one authorized execution twin. The policy applies to every prompt whose frontmatter d→ GUIDEProject Cleanup and Uninstall GuideThe 2.0 uninstaller removes only a validated Mission Directives project installation and its managed guidance and ignore blocks. It does not infer ownership from common directory n→ GUIDEProject Config Guide`.mission-directives/project.json` is durable, agent-facing project knowledge. `.mission-directives/config.json` is system-owned installation and viewer configuration. Project Conf→ GUIDEPrompt Addition and Registration Guide**Purpose:** Add one new prompt to Mission Directives from a Markdown source file without manually editing every catalog, graph, fixture, routing table, or integrity receipt. **Can→ GUIDEPrompt Body Authoring GuideThis manual explains how to create or revise a prompt body that is precise, composable, least-privileged, machine-parseable, and verifiable. It supplements the normative `PROMPT_ST→ GUIDEPrompt Body Validation GuideThe suite validates prompt bodies separately from frontmatter, catalogs, scenarios, schemas, skills, and runtime state. This prevents a structurally complete catalog from hiding we→ GUIDERecovery and Rollback GuideRecovery is a designed and verified capability, not a sentence in an action plan. This guide explains how consequential runs prepare, exercise, select, and verify rollback, roll-fo→ GUIDERoot Agent Guidance and MD Keyword Routing GuideThis guide explains how to connect a project repository's root agent-instruction files to the Mission Directives without copying prompt bodies into those files or overwriting human→ GUIDERouter Keyword Catalog and Scoring GuideThis guide defines the deterministic intent-routing vocabulary, scoring model, confidence behavior, diagnostics, and extension workflow. The runtime source of truth is `config/rout→ GUIDEScenario Authoring GuideA scenario is a reusable cross-capability execution route. It is justified when the same combination of evidence, planning, production, review, authorization, and verification recu→ GUIDESecurity Testing and Simulation BoundariesSecurity prompts are defensive capabilities for systems the operator owns or is explicitly authorized to test.→ GUIDESecurity Operations GuideThis guide applies to defensive security review, authorized simulations, exploit reproduction, model and agent security, incident response, OSINT, and security remediation.→ GUIDESkill Installation and Integration GuideInstall **only the exact skill needed**. Every command below uses the per-skill form; it does not install the full repository. Review the current source and audit results before in→ GUIDESkill RoutingSkills are optional execution adapters, not hidden prompt powers. The router probes availability, trust, and fit before selection; unavailable skills never block the native prompt → GUIDESkill Supply-Chain GuideThird-party skills can improve specialized work, but they are executable supply-chain dependencies. This guide defines how skills are discovered, locked, reviewed, tested, quaranti→ GUIDETelemetry and Observability GuideTelemetry makes routing, execution, verification, failure, and improvement observable without storing unnecessary sensitive content. It supports model selection, cost control, drif→ GUIDETemplate System Guide[Back to Mastery Manual](MD_MASTERY_MANUAL.md) · [Manual index](MANUALS.md)→ GUIDETool Policy and Authorization GuideAuthorization and tool selection are related but different:→ GUIDETroubleshooting GuideFor the complete GitHub Actions incident history, fixing commits, and required prevention checklist, see [GitHub Actions Failure History and Pre-Push Guide](GITHUB_ACTIONS_FAILURE_→ GUIDEVisual Assets Integration Guide`visual-assets` is a user-owned local skill for deliberate, editorially designed visual construction. It builds assets from primitives:→