Skip to content

Skills, Agents & Hooks

Skills are where ido4’s project intelligence lives. Each skill composes multiple MCP tools into an intelligent workflow — assembling context, spotting patterns, and delivering actionable insights. Every claim is backed by real data from the audit trail and analytics. Nothing is guessed.

Together with 4 specialized agents and 2 automation hooks, the plugin layer turns raw governance data into the project understanding that makes AI-hybrid development work.

Complete Inventory

Project Intelligence (any methodology)

ComponentTypeWhat it does
/ido4dev:standupSkillMorning briefing — what’s blocked, what’s at risk, the single highest-leverage action for the day
/ido4dev:boardSkillFlow intelligence — cascade blockers, false statuses, review bottlenecks, epic cohesion
/ido4dev:healthSkillGREEN / YELLOW / RED verdict across flow, governance, and team dimensions
/ido4dev:complianceSkillQuantitative score (0-100) + structural principle audit + improvement recommendations

Planning (methodology-specific)

ComponentTypeMethodologyWhat it produces
/ido4dev:plan-waveSkillHydroValid-by-construction wave plan respecting all 5 governance principles
/ido4dev:plan-sprintSkillScrumSprint backlog with Definition of Ready gates per work item type
/ido4dev:plan-cycleSkillShape UpBetting table with appetite check and circuit breaker risk assessment

Retrospectives (methodology-specific)

ComponentTypeMethodologyWhat it analyzes
/ido4dev:retro-waveSkillHydroVelocity, epic integrity, blocking time — real data from audit trail
/ido4dev:retro-sprintSkillScrumSprint goal achievement, DoR effectiveness, carry-over trends
/ido4dev:retro-cycleSkillShape UpBet outcomes, appetite accuracy, circuit breaker decisions

Specification & Decomposition

ComponentTypeWhat it does
/ido4dev:decomposeSkillTransforms a strategic spec into a technical spec with implementation tasks grounded in your codebase
/ido4dev:spec-validateSkillCatches format and quality issues before ingestion
/ido4dev:spec-qualitySkillQuality standards for task descriptions, success conditions, effort/risk calibration
Code AnalyzerAgentMaps strategic capabilities to codebase modules, discovers patterns. Uses Read/Glob/Grep. Model: Opus
Technical Spec WriterAgentDecomposes capabilities into right-sized tasks with code-grounded metadata. Model: Opus
Spec ReviewerAgentIndependent two-stage review — format compliance + quality assessment. Model: Sonnet

Sandbox & Onboarding

ComponentTypeWhat it does
/ido4dev:onboardSkillZero-friction first touch — auto-clones demo codebase, creates sandbox, guided governance discovery
/ido4dev:guided-demoSkillFour-act governance walkthrough — project overview, violation discovery, enforcement, full pipeline
/ido4dev:sandbox-exploreSkillInteractive exploration — 13 structured paths across governance, enforcement, coordination
/ido4dev:sandboxSkillSandbox lifecycle management — create, reset, destroy
/ido4dev:pilot-testSkillEnd-to-end verification that the full governance stack works

Persistent Intelligence

ComponentTypeWhat it does
PM AgentAgentPersistent project intelligence brain. Maintains velocity baselines, compliance trends, blocker patterns across sessions. Grounds every recommendation in real data. Model: Sonnet
Post-transitionHookFires after state changes. Checks: did this unblock downstream? Create new blocker? Reach milestone?
Post-wave-assignmentHookFires after wave assignment. Checks epic integrity and dependency violations.

Totals: 21 skills + 4 agents + 2 hooks

How Skills Work

Skills are SKILL.md files — structured prompts that tell the AI what to do:

  1. Call specific MCP tools to gather data
  2. Analyze results for patterns and anomalies
  3. Present findings in a consistent format
  4. Recommend actions grounded in evidence

Composite Data Tools (the performance layer)

Skills don’t make 10+ individual tool calls. They call composite aggregators that assemble everything in one request:

AggregatorWhat it assemblesUsed by
get_standup_dataContainer status, tasks, PR reviews, blocker analyses, audit trail (24h), analytics, agents, compliance/ido4dev:standup
get_board_dataContainer status, tasks with PR + lock annotations, analytics, agents/ido4dev:board
get_compliance_dataCompliance score, audit trail, analytics, tasks, blocker analyses, integrity checks/ido4dev:compliance
get_health_dataContainer status, compliance, analytics, agents/ido4dev:health
get_task_execution_dataTask spec, upstream context (what was built), sibling patterns, downstream consumers, epic progress, risk flagsTask execution

These are MCP tools defined in @ido4/mcp — available to any MCP client, not just skills.

Skills vs Prompts vs Agents

SkillsPromptsAgents
PlatformClaude Code (with plugin)Any MCP clientClaude Code (with plugin)
Invocation/ido4dev:standupMCP prompt protocolInvoked by skills
FeaturesMemory, file access, hooksTool calls + reasoningSpecialized instructions, model selection
Count2184
PurposeIntelligent workflowsPortable guidanceFocused AI roles

Cross-Skill Intelligence Loop

Skills share knowledge through Claude Code’s memory, creating a compounding intelligence loop:

  • Retro analyzes what happened → persists findings (velocity baselines, blocking patterns, process gaps)
  • Standup reads retro findings → contextualizes today’s risks against historical patterns
  • Plan reads velocity baselines → grounds capacity estimates in real throughput, not guesses
  • Compliance tracks score trends → detects governance degradation across assessments
  • PM Agent bridges everything → persistent memory across sessions, coordination across agents

Every skill invocation makes the next one smarter. This is institutional memory in action.

PM Agent

The PM agent is a persistent project intelligence brain — not a chatbot, but a data-grounded advisor:

  • Maintains velocity baselines, compliance trends, and blocker patterns across sessions
  • Grounds every recommendation in real data — audit trail, analytics, compliance score
  • Coordinates multiple agents — detects lock contention, idle agents, work imbalance
  • Cannot override the BRE — translates validation failures into actionable guidance

See PM Agent for details.