process skills
61 skills. Source of truth: skills/process/.
adversarial-review
Isolated, cross-model adversarial review of KBD artifacts and change diffs. Dispatches a fresh-context LLM judge over an OpenAI-compatible REST gateway (openai-proxy or liter-llm api) with an explicit mandate to find problems — the model that produced an artifact or change is never the model that reviews it. Runs as a pipeline stage inside kbd-assess/analyze/plan (artifact mode) and kbd-execute's per-change QA gate (diff mode). Findings are severity-bucketed (CRITICAL/WARNING/SUGGESTION) and the judge's own report is screened by sycophancy-correction before it is surfaced.
v1.0.0 · process review quality llm-as-judge kbd · source
agent-team-creator
Create the smallest useful coding agent team for a task, with guided role discovery and native exports for UAR, Codex, Claude Code, Copilot, Kimi Code, MiniMax CLI, OpenCode and DeepSeek. Use when a user asks to create, configure or choose an agent team; use agent-team-manage for existing task state. Do not use for existing task updates (see agent-team-manage).
v1.0.0 · agents teams orchestration coding · source
agent-team-handoff
Prepare and accept a durable handoff between agent roles or coding harnesses using task context, Git identity, evidence and memory references. Use when work must continue in another harness or agent; use agent-team-manage for ordinary same-owner task updates. Do not use for ordinary same-owner task updates (see agent-team-manage).
v1.0.0 · agents teams orchestration coding · source
agent-team-manage
Manage an existing agent team: assign tasks, dependencies and owners, update definitions, track evidence, cancel or reassign work, and reconcile KBD-linked completion. Use when managing team lifecycle requests; use agent-team-creator for initial team discovery and agent-team-handoff to transfer work across harnesses. Do not use for initial team creation (see agent-team-creator).
v1.0.0 · agents teams orchestration coding · source
agent-team-models
Discover configured models and choose explicit capability, strength and cost policies at team, role, skill and task levels. Use when selecting models or controlling an agent team budget; use agent-team-creator for role discovery. Unknown prices and capabilities remain unknown. Do not use for role selection (see agent-team-creator).
v1.0.0 · agents teams orchestration coding · source
clone-skill (sub-skill)
Clone an existing Agent Skill and adapt it for a new domain. Preserves architecture while replacing domain-specific content.
v1.0.0 · process orchestration automation · source
cowork-management
Install, update, and manage AI coding agent skills across 20+ platforms using the cowork CLI. Covers Claude Code, Codex, OpenCode, Kimi Code, Zed, Cursor, Windsurf, GitHub Copilot, and more. Also manages prometheus-skill-pack updates, toolchain health checks, and disk-space reclamation via dsg delegation.
v1.0.0 · cli skill-management cowork install platform toolchain prometheus · source
create-native-agent (sub-skill)
Generate a complete native Rust agent application with Supabase-style CLI, A2A/AG-UI/A2UI protocol support, React 19 assistant-ui chat frontend, liter-llm model routing, MCP client tools, and the Prometheus skills engine. One command scaffolds a runnable multi-protocol agent binary.
v1.0.0 · process orchestration automation · source
create-skill (sub-skill)
Create a new Agent Skill from scratch using PMPO orchestration. Generates a complete skill package with all required files.
v1.0.0 · process orchestration automation · source
evolve (sub-skill)
Run a full iterative evolution cycle — Assess, Analyze, Plan, Execute, and Reflect. Auto-detects domain or prompts for it. Use for any subject that needs iterative improvement against goals.
v1.0.0 · process orchestration automation · source
evolve-analyze (sub-skill)
Run only the Analyze phase — research the external landscape for competitors, trends, opportunities, and threats. Requires an existing assessment or goals.
v1.0.0 · process orchestration automation · source
evolve-assess (sub-skill)
Run only the Assess phase — evaluate current state against goals. Produces a structured assessment without creating plans or making changes.
v1.0.0 · process orchestration automation · source
evolve-execute (sub-skill)
Run only the Execute phase — carry out a previously created improvement plan. For software domain evolutions, delegates to kbd-process-orchestrator as the inner loop for fine-grained task execution with OpenSpec detection and artifact-refiner QA. Requires an existing plan.json.
v1.0.0 · process orchestration automation · source
evolve-plan (sub-skill)
Run only the Plan phase — create a prioritized improvement plan. Requires an existing assessment and optionally an analysis.
v1.0.0 · process orchestration automation · source
evolve-report (sub-skill)
Generate an evolution report from current state data. Produces a human-readable markdown report summarizing the full evolution cycle. Aggregates KBD phase results and artifact-refiner QA metrics when available. Optionally refines the report through artifact-refiner content refinement.
v1.0.0 · process orchestration automation · source
evolve-status (sub-skill)
Check the current status of an evolution cycle. Shows iteration count, goal alignment, phase completion, and convergence status.
v1.0.0 · process orchestration automation · source
extend-skill (sub-skill)
Add new capabilities to an existing Agent Skill without breaking current functionality. Non-destructive additions only.
v1.0.0 · process orchestration automation · source
ideation-mindmap
Stage-zero onramp for /start-business-build. Takes a one-line business concept and generates a 6-branch concept mindmap via surreal-memory, structuring raw ideas into actionable branches ready for zeespec constraint capture.
v1.0.0 · ideation mindmap surreal-memory business-build stage-zero · source
iterative-evolver
Domain-agnostic PMPO improvement loop: Assess -> Analyze -> Plan -> Execute -> Reflect over any subject (software, product, business, research, operations). Use when the user wants to improve or evolve something over repeated passes, asks "make this better", "iterate on this", "evolve this", or when a target needs assessing before changes are planned. Do NOT use for a single scoped code change (just make it), for standing always-on automation with termination criteria (see pmpo-outer-loop), or for gathering a missing fact mid-stage (see pmpo-elicit).
v1.0.0 · process orchestration automation · source
karpathy-progress-memory
Record a successful KBD task, change, or phase boundary as a canonical, idempotent progress event. Use when a boundary transition succeeds, when recovering work after context loss, or when KBD progress must survive a memory-service outage without editing generated projections. Do NOT use for changing KBD position or status; use kbd-process-orchestrator instead.
v1.0.0 · process kbd progress memory recovery · source
kbd-analyze (sub-skill)
Run the Analyze stage of the KBD lifecycle: research the engineering landscape — existing open-source libraries, frameworks, and skeletons that fit the assessed gaps, plus stack discovery when none is specified — and write the candidate set that the Spec and Plan stages consume.
v1.0.0 · process orchestration automation · source
kbd-apply (sub-skill)
KBD-owned spec-apply driver. Wraps a spec backend (OpenSpec; Spec Kit via the speckit adapter) and drives it ONE task at a time, so KBD stays the source of truth: every task boundary fires KBD hooks, emits a plain-text position signal, and syncs progress.json and the waypoint. Replaces the broken pattern of handing the turn to a bare /opsx:apply that runs outside KBD.
v1.0.0 · process orchestration automation openspec spec-kit · source
kbd-assess (sub-skill)
Assess the current project codebase against the active phase goals. Project-agnostic: reads AGENTS.md, spec files, and codebase to produce a structured assessment. Works for any technology stack or language.
v1.0.0 · process orchestration automation · source
kbd-audit (sub-skill)
Audit the exact KBD position, lifecycle history, plan revision, ownership, blockers, and uncommitted work without mutating the run.
v1.0.0 · process orchestration control audit · source
kbd-bottleneck-detector (sub-skill)
Evaluate or repair canonical KBD task, phase, and ZeeSpec boundaries. Use when progress receipts, projections, or build gates may be stale, or when the user mentions "bottleneck detector". Do NOT use for creating or advancing phases (see kbd-new-child and kbd-next-phase).
v1.0.0 · process orchestration recovery · source
kbd-cancel (sub-skill)
Gracefully cancel the active KBD run while preserving its checkpoint and immutable audit history.
v1.0.0 · process orchestration control cancel · source
kbd-child-exit (sub-skill)
Exit the active KBD child loop: write its handoff-out, roll its progress up to the parent node, pop the position path, and return control to the parent. The --enter companion descends into a selected child so new children nest under it.
v1.0.0 · process orchestration automation · source
kbd-doctor (sub-skill)
Diagnose Prometheus substrate health and orchestrate safe doctor runs through the compiled prometheus doctor CLI. Use when the user asks for doctor, health checks, repair planning, refresh planning, or recovery of KBD learning-loop infrastructure.
v1.0.0 · kbd doctor repair refresh health recovery cli · source
kbd-evolve
Domain-landscape-first evolution for KBD projects. Unlike /kbd-next-phase (which advances the next planned phase), /kbd-evolve surveys the external landscape of the project's problem domain, identifies the highest-impact improvement opportunities against configurable criteria, and produces a ranked evolution brief that /kbd-new-phase or /kbd-process-orchestrator can consume as the seed for the next phase.
v1.0.0 · process orchestration evolution research landscape kbd pmpo · source
kbd-execute (sub-skill)
Select an execution backend for the active KBD phase, write canonical phase execution state, dispatch the phase to the appropriate tool or OpenSpec, and maintain KBD as the source of truth. Supports multi-tool handoff via progress.json protocol. Runs integration and review gates once at phase completion.
v1.0.0 · process orchestration automation · source
kbd-goal
Platform-agnostic goal-driven iterative loop. State a goal; KBD orchestrates Ideation → Specification → Creation phases autonomously until the goal is met, a budget is exhausted, or human approval is needed. Works identically on Claude Code, Codex CLI, OpenCode, Kimi Code, and Zed — routing to each platform's native /goal primitive where available, implementing the evaluator pattern where not.
v1.0.0 · process orchestration goal autonomous loop multi-platform · source
kbd-goal-check
Kimi Code goal-condition evaluator skill. After each Kimi turn, checks whether the active phase's stopping condition has been met. Returns PASS (with evidence) or CONTINUE (with next action from TASKS.md). Implements the maker≠evaluator pattern on Kimi Code, where /goal next is a queue (not a condition-based loop) and requires an external evaluator.
v1.0.0 · process goal evaluator kimi loop condition-check · source
kbd-init (sub-skill)
Initialize KBD in a project. Auto-discovers project identity, stack, and constraints from existing context files (AGENTS.md, CLAUDE.md, README.md, package.json, Cargo.toml, pyproject.toml, etc.) and generates .kbd-orchestrator/project.json and .kbd-orchestrator/constraints.md. Run this once per project before any other KBD command.
v1.1.0 · process orchestration automation · source
kbd-inject-agent-rules (sub-skill)
Idempotently inject the agent-rules or UI/UX-routing managed pack into a target project's CLAUDE.md and/or AGENTS.md. Re-runnable — overwrites only the selected fenced region; everything else is byte-preserved. Supports --refresh to re-validate cached source URLs, and --dry-run for a target-file diff preview.
v1.0.0 · process orchestration docs agent-rules · source
kbd-memory-recall (sub-skill)
Query surreal-memory for prior similar KBD work and write a markdown digest at .kbd-orchestrator/phases/<phase>/prior-context.md. Used as planning input before /kbd-assess. Auto-invoked via the assess:before hook; degrades gracefully when the memory endpoint is unreachable.
v1.0.0 · process orchestration memory learning · source
kbd-new-child (sub-skill)
Create a child phase inside the currently-active top-level phase. Mirrors /kbd-new-phase but writes into phases/<parent>/children/<child>/, appends the new child to childPhases[], moves childPointer to it, and fires child:before. Use to split a parent phase into scoped sub-processes.
v1.0.0 · process orchestration automation nested-phases · source
kbd-new-phase (sub-skill)
Manually create a new top-level KBD phase. Accepts <name> [goals…] and initialises the phase directory, waypoint, project.json activePhase, and fires phase:before. Use this when no prior reflection exists (e.g. very first phase of a new project), when pivoting away from /kbd-next-phase's suggestion, or when initialising state by hand.
v1.0.0 · process orchestration automation · source
kbd-next-child (sub-skill)
Advance childPointer to the next entry in childPhases[], or jump to a named child directly. Fires child:after for the closing child and child:before for the new active child. Refuses to advance past the last child (suggesting /kbd-reflect + /kbd-next-phase instead).
v1.0.0 · process orchestration automation nested-phases · source
kbd-next-phase (sub-skill)
Continue to the next KBD phase, automatically seeded from the previous phase's reflection. Reads the "Recommended Next Phase" section of reflection.md, initializes the new phase directory with goals.md and a skeleton progress.json, updates current-waypoint.json to point to /kbd-assess, and updates project.json activePhase.
v1.0.0 · process orchestration automation · source
kbd-pause (sub-skill)
Gracefully pause the active KBD run, checkpoint its exact position, and prevent every harness from steering execution until an operator resumes it.
v1.0.0 · process orchestration control pause · source
kbd-plan (sub-skill)
Create a prioritized, ordered change list for the current project phase. Project-agnostic: reads assessment and project constraints to produce an ordered change list. Auto-detects OpenSpec availability and emits appropriate change formats. Bridges with iterative-evolver plan items when running inside an evolution cycle.
v1.0.0 · process orchestration automation · source
kbd-process-orchestrator
The Universal Knowledge-Based Development (KBD) process orchestrator. Drives the full iterative PMPO lifecycle for ANY project — Assess → Analyze → Plan → Execute (backend dispatch) → Reflect — at every granularity level: global phases, OpenSpec changes, and artifact-level QA. Implements the process defined in TJ-KBD-UNIVERSAL-001. Coordinates execution across multiple AI tools (Claude Code, Codex, OpenCode, Kimi, and compatible adapters) using the append-only KBD runtime as the shared control plane.
v2.0.0 · process orchestration automation · source
kbd-reflect (sub-skill)
Generate the phase reflection report after all changes in the phase are complete. Seeds the next phase's planning inputs. Aggregates artifact-refiner QA results and reports back to iterative-evolver when running inside an evolution cycle.
v1.0.0 · process orchestration automation · source
kbd-resume (sub-skill)
Resume a paused KBD run after validating its checkpoint and plan revision.
v1.0.0 · process orchestration control resume · source
kbd-spec (sub-skill)
Run the Spec stage of the KBD lifecycle: turn an assessment and analysis into concrete, ordered changes — native-kbd specs (spec.md + tasks.json + verification.md) or OpenSpec proposals — gated by ZeeSpec requirements coverage when present.
v1.0.0 · process orchestration automation · source
kbd-status (sub-skill)
Show current KBD process status — active phase, change inventory, goal completion, and next recommended action. Reads progress.json to surface work completed by all tools (Antigravity, Roo, Cursor, Cline, Codex, etc.)
v1.0.0 · process orchestration automation · source
liter-llm-bridge
Harness-agnostic multi-model routing fallback. Builds and configures liter-llm (a Rust LLM proxy with a built-in MCP server) so any harness with MCP support can dispatch each phase of KBD / iterative-evolver / artifact-refiner to the cheapest model that meets the cognitive requirement. Use when the host harness (Claude Code, codex, opencode, cursor, etc.) does not natively support per-skill model selection.
v0.1.0 · process orchestration automation · source
native-agent
Generates a complete, production-ready native Rust agent application with a Supabase-style management CLI. The generated agent binary embeds an Axum HTTP server supporting A2A, AG-UI, and A2UI protocols, a React 19 assistant-ui chat frontend, liter-llm provider routing, simple MCP client tool integration, the Prometheus skill pack selection engine, and full Docker support with Docker Desktop auto-detection and image loading. One command scaffolds a runnable, configurable, containerizable multi-protocol agent.
v1.0.0 · process orchestration automation · source
pmpo-elicit
Ask-or-research primitive that resolves a missing fact mid-lifecycle and records it with provenance instead of guessing. Use when a stage is blocked on unknown information, when the user says "I don't know, find out", or when a decision needs a fact nobody has supplied. Do NOT use to run an improvement loop (see iterative-evolver) or to define standing automation (see pmpo-outer-loop).
v1.0.0 · process orchestration automation · source
pmpo-evolver
Strategy router for evolving released products — routes to one of five evolution perspectives (competitive, trend, unique-product, idea-validation, self-learning) and orchestrates the full PMPO loop per perspective
v1.0.0 · pmpo evolution strategy competitive-analysis self-learning liter-llm iterative-evolver · source
pmpo-outer-loop
Defines a standing outer loop — goal, feedback sources, termination criteria — that repeatedly drives PMPO/KBD cycles and only escalates at declared decision points. Use when the user wants work to keep running unattended, mentions "outer loop", "standing loop", "run until", "autonomous loop", or asks to be interrupted only at specific checkpoints. Do NOT use for one bounded improvement pass (see iterative-evolver) or for the tactical per-change KBD cycle (see kbd-process-orchestrator).
v1.0.0 · process orchestration automation · source
pmpo-skill-creator
Use this skill to create, clone, or extend Agent Skills using PMPO orchestration. Generates complete, production-ready skills with agentskills.io spec compliance, Claude Code plugin/marketplace support, OpenCode tools compatibility, state management, lifecycle hooks, and workflow triggers. Supports three modes: create (from scratch), clone (adapt existing skill to new domain), and extend (add capabilities to existing skill).
v1.0.0 · process orchestration automation · source
prometheus-context-bootstrap
Scaffold the Prometheus agent structure into a new or existing project — a portable AGENTS.md carrying compaction-surviving invariants, CLAUDE.md pointing at it, path-scoped rules per detected stack, deterministic hooks for phase-gated integration verification and single-writer builds, .prometheus append-only learning, and a skill-budget-safe settings.json. Profile-aware for mixed model fleets: includes an execution scaffold by default for non-frontier models, omits it only when measured. Creates what is absent, splices a marked region into what exists, and never overwrites operator prose.
v1.0.0 · bootstrap context-engineering agents-md claude-md hooks scaffold · source
start-business-build (sub-skill)
Top-level orchestrator that takes a one-line business concept ("track competitor pricing", "summarize support tickets nightly") and chains every Prometheus pipeline stage end-to-end — ideation expansion, zeespec constraint capture, iterative-evolver assess+plan, OpenSpec change generation, forge enrich, AI implementation dispatch, forge reflect, pk ingest — and optionally finishes by packaging the result as a LibreFang WASM skill and offering /upload-to-bossfang. Single command from concept to deployable skill.
v1.0.0 · orchestrator pipeline ideation headline end-to-end · source
upload-to-bossfang (sub-skill)
Upload a packaged .lf-skill.zip to a running LibreFang/bossfang instance via its /skills/install REST endpoint. Validates the URL with a deny-by-default SSRF guard (rejects private IPs, non-http schemes, and unallowlisted hosts unless --insecure is passed for explicit local-dev use). Auto-discovers the zip in cwd, redacts the BOSSFANG_TOKEN bearer auth from all logs, and verifies the install by GETting the skill manifest after upload. Use after forge package-librefang produces a zip and you want to deploy it to a remote bossfang.
v1.0.0 · librefang bossfang deploy upload ssrf-guard security-reviewed · source
validate-idea (sub-skill)
Staged idea validation for product evolution — three gates from plausibility to specification using the Darwin Gödel Machine pattern, with Archive of Stepping Stones persistence
v1.0.0 · idea-validation evolution darwin staged-gating liter-llm pmpo-evolver · source
validate-skill (sub-skill)
Validate any Agent Skill against the agentskills.io specification and quality standards. Reports compliance issues and suggestions.
v1.0.0 · process orchestration automation · source
zeespec-interrogate (sub-skill)
Run a full ZeeSpec 5W1H interrogation on a named subject. Asks 10 questions per dimension (Why, Who, When, What, Where, How), scores coverage, and produces a constraint manifest with a GO/CAUTION/NO-GO recommendation.
v1.0.0 · process orchestration automation · source
zeespec-interrogator
Zachman Framework-based specification interrogation skill. Applies the ZeeSpec 5W1H method (What, Where, Who, When, Why, How) across 6 dimensions x 10 questions to surface undefined system constraints before planning or implementation begins. Invoked standalone for ideation-layer GO/NO-GO decisions, or triggered automatically by kbd-process-orchestrator and iterative-evolver when spec coverage falls below threshold. Produces a structured constraint manifest consumed by the calling process.
v1.0.0 · process orchestration automation · source
zeespec-score (sub-skill)
Score and report coverage on an existing ZeeSpec interrogation record. Reads the recorded answers from state and produces a coverage-score.json with per-dimension scores and a GO/CAUTION/NO-GO recommendation without re-running the interrogation.
v1.0.0 · process orchestration automation · source
zeespec-status (sub-skill)
Show current ZeeSpec interrogation progress for a named subject. Displays which dimensions have been completed, current coverage scores if available, and the manifest path if the interrogation is complete. Read-only — does not modify state.
v1.0.0 · process orchestration automation · source