Homework validator — checks that user completed the QUICKSTART tutorial correctly. Runs 5 SQL/file checks (routing.db populated, project registered, demo workspace, skill invocations logged, task folder structure) plus an optional 6th self-audit step. Reports ✅/❌ per check with remediation hints. Use when: after completing QUICKSTART § 5 walkthrough, "/tutorial-check", "check my setup", "проверь мою установку", "run homework checks".
Protocol for architecturally clean execution of complex multi-block tasks. Decomposition into blocks → research before library-first → explicit decision pinning → clear scope vs out-of-scope → report after each block. Principle: every architectural decision is pinned explicitly, no ambiguity, epics don't sprawl. Use when: task has 5+ blocks; DB migration; rewriting skills; any task where architectural cleanliness must be preserved over distance
Automatic linking of an artifact to project architecture elements. Two modes: file-mode (by file paths, for fast-track flows) and feature-mode (by artifact text, for briefs/specs). Merge logic: appends to existing arch_ref, does not overwrite. Propagate: final arch_ref written up the chain epic → spec → brief. Use when: after plan-first in fast-track / pipeline; after saving brief / spec artifact
Audit-before-fix protocol — first find all gaps across 7 planes, prioritize, pin the gap table, and only then fix. Principle: never touch code until the full picture of problems is understood. Gap table → approval → flow-first per gap → library-first → plan → code. Use when: task begins with looking for problems ("find gaps in X", "audit security", "what's wrong with Y"), not with a specific fix
Meta-orchestrator: assigns skill-chain cadence per block (Tier 1/2/3) via Q1-Q6 rules. Reads target task_blocks + task.md, writes cadence-decisions-{R}.md artifact. Standalone (executor invokes) or Batch (generator hand-off). Use when: deciding cadence for a batch of blocks, generator hand-off from decomposition skills
Coverage validation protocol before a final approval gate. Reads requirements from any structured requirements document, matches them against the proposed decomposition units (blocks, epics, tasks), outputs a coverage table. If gaps exist — stops, waits for user decision. If everything is covered — continues without pause. Use when: after proposing any decomposition of work (blocks / epics / tasks) before committing to execution — as a safety net that no requirement fell through the cracks.
Makes an architectural / project / scope decision using a 5-part model INSTEAD of asking the user. Structure: 🎯 Decision / Why / 🛡 Security / 📈 Scalability / Alternatives / Plain-language analogy. 1 question = 1 atomic artifact. Use when: the agent is about to ask an architectural / scope question; "/decision-first", "decide it yourself", "apply the decision model"
Autonomous orchestrator of the per-block go-fast cycle. Takes over after arch-first/audit-first when "autopilot" mode is chosen. Runs every block: flow-first → library-first → plan-first → execute without manual approvals. Validates each skill's artifact against built-in rules. On blockers — escalates to a chat bot and waits for a reply. Use when: arch-first or audit-first asks "how do we work?" and the user chooses autopilot mode; "/dev-auto-first", "autopilot", "run automatically"