GxFn

Submit new Alembic knowledge candidates from Codex. Covers `alembic_submit_knowledge`, required V3 fields, batch rules, and review boundaries.

Generate Alembic Wiki documentation through `alembic_wiki` using plan, article writing, and finalize.

Check code against Alembic Recipe standards with `alembic_guard`. Trigger for audits, lint-like review, diff checks, compliance checks, and "does this follow our conventions?" questions.

Use Alembic Recipes as the project's curated source of truth for conventions, code patterns, facts, and Guard standards. Trigger when answering questions about project standards or when implementation should follow local knowledge.

Use Alembic from Codex. Start here for status, initialization, project priming, Guard checks, bootstrap/rescan jobs, Dashboard handoff, and permission boundaries.

Discover Alembic project structure, targets, files, metadata, dependency relationships, and Recipe graph context.

Use when: generating source-derived chain maps and long-chain execution plans, validating long-chain workflow behavior node by node, reviewing benchmark coverage, applying optional domain overlays, or repairing Alembic cold-start/rescan/bootstrap/delivery/skill-generation flows with explicit evidence and write boundaries.

Submit knowledge to Alembic. Covers single/batch MCP submission, V3 field requirements, quality validation, and lifecycle. Use when user says "submit knowledge / add to KB / create recipe" or agent needs to persist code patterns, rules, or facts.

Use in a Wakeflow Design window to turn a clarified and confirmed requirement into a controller-intake-ready requirement design with user stories, decisions, tests, non-goals, and acceptance criteria.

Use in a Wakeflow Test window to design a risk-based validation plan for a controller test card, real-project scenario, release check, runtime observation, or cross-repository integration question.

Use in a Wakeflow Design window to prepare a compact handoff for controller intake from clarified requirements, option plans, requirement designs, and Design signals.

Use in a Wakeflow Design window when a clarified requirement needs multiple implementation, architecture, sequencing, or rollout options before controller or user confirmation.

Use in a Wakeflow Design window when a user idea, controller request, handoff candidate, or product change needs clarification before it can become an original plan, requirement design, TODO candidate, or controller decision.

Use in a Wakeflow Design window to break a confirmed requirement design into vertical-slice TODO or task-package candidates for controller intake.

Use when Test needs to run or monitor reusable Alembic test routes: environment readiness, BiliDili/AlembicWorkspace cold-start, resident service, multi-root ProjectScope, or explicitly assigned Codex Plugin / host MCP evidence. Follow verified routes and collect evidence instead of inventing new paths.

Use in a Wakeflow Test window when a test card, smoke run, runtime observation, or user report shows a failing or uncertain behavior that needs reproduction, root-cause tracing, and ownership classification.

Use when a Wakeflow development/product window is implementing its assigned task package and needs execution craft — plan-first, clean baseline, test-first, systematic debugging, two-stage self-review, scope discipline, and verify-before-done. It shapes HOW the code is written so the window naturally produces the evidence the controller's acceptance gate requires, and how to receive a rework verdict without blind re-patching. It never changes the delivery/return protocol (that stays with wakeflow-target) and grants no claim, accept, dispatch, cross-window, or state-machine-write authority.

Use when a tmux-resident target Claude Code window session receives a Wakeflow Delivery Loop direct-thread delivery in its tmux pane, executes only its assigned dispatch packet, reports a TargetResultEnvelope, sends the controller-return with the wakeflow-claude-host helper, or enforces target-window boundaries without claim / finish / chain-next state.

Use when a target Codex window receives a Wakeflow Delivery Loop direct-thread delivery, executes only its assigned dispatch packet, reports a TargetResultEnvelope, or enforces target-window boundaries without claim / finish / chain-next state.

Use when Wakeflow total control starts or resumes Wakeflow Delivery Loop in Claude Code, reviews target result envelopes, creates dispatch packets, builds delivery envelopes, sends deliveries to tmux-resident window sessions with the wakeflow-claude-host helper, decides acceptance / rework / block / next wave, or stops unattended automation.

Use when working inside Wakeflow on workspace initialization / setup, CLAUDE.md / skill layering, TODO / Backlog intake, Design handoff intake, idle-window scheduling, window coverage, task-package dispatch, producer/consumer sequencing, unified dispatch prompts, test handoffs, validation boundaries, or workspace script pipelines. This skill supplements CLAUDE.md and must not override its hard boundaries.

Use when Wakeflow total control starts or resumes Wakeflow Delivery Loop, reviews target result envelopes, creates dispatch packets, builds delivery envelopes, decides acceptance / rework / block / next wave, or stops unattended automation.

Use when working inside Wakeflow on workspace initialization / setup, AGENTS.md / skill layering, TODO / Backlog intake, Design handoff intake, idle-window scheduling, window coverage, task-package dispatch, producer/consumer sequencing, unified dispatch prompts, test handoffs, validation boundaries, or workspace script pipelines. This skill supplements AGENTS.md and must not override its hard boundaries.

Submit knowledge to Alembic. Covers single/batch MCP submission, V3 field requirements, quality validation, and lifecycle. Use when user says "submit knowledge / add to KB / create recipe" or agent needs to persist code patterns, rules, or facts.

Generate and browse project Wiki documentation through the Alembic resident daemon's wiki HTTP API (generate → status → files). Use when user says "generate wiki/docs", "write documentation", or agent needs structured project documentation produced from the knowledge base.

Guard checks code against project Recipe standards via MCP tool alembic_guard (auto-routes by code/files params). Use when the user wants to audit, lint, or verify code compliance.

Provides this project's Recipe-based context to the agent. Recipes are the project's standard knowledge (code patterns + usage guides + structured relations). Use when answering about project standards, Guard, conventions, or when suggesting code. Supports in-context lookup, terminal search (alembic search), and on-demand semantic search via MCP tool alembic_search (mode=context).

Discover project structure (targets, files, dependency graph) and browse the knowledge graph (relations between Recipes). Use when the user asks about module structure, project targets, dependency relationships, or knowledge graph navigation.

Submit knowledge to Alembic. Covers single/batch MCP submission, V3 field requirements, quality validation, and lifecycle. Use when user says "submit knowledge / add to KB / create recipe" or agent needs to persist code patterns, rules, or facts.

Generate and publish project Wiki documentation using asd_wiki MCP tool (plan → write → finalize). Use when user says "generate wiki/docs", "write documentation", or agent needs to produce structured project documentation from the knowledge base.

Guard checks code against project Recipe standards via MCP tool asd_guard (auto-routes by code/files params). Use when the user wants to audit, lint, or verify code compliance.

Provides this project's Recipe-based context to the agent. Recipes are the project's standard knowledge (code patterns + usage guides + structured relations). Use when answering about project standards, Guard, conventions, or when suggesting code. Supports in-context lookup, terminal search (asd search), and on-demand semantic search via MCP tool asd_search (mode=context).

Discover project structure (targets, files, dependency graph) and browse the knowledge graph (relations between Recipes). Use when the user asks about module structure, project targets, dependency relationships, or knowledge graph navigation.

Use Alembic Recipes proactively only when the current project has a project-level Alembic knowledge skill or local Alembic knowledge base. For empty projects, use only when the user explicitly asks for Alembic Recipes or setup/status.

Use Alembic setup/status/diagnostics/bootstrap/rescan when the user explicitly asks for Alembic. When the current project has a project-level Alembic knowledge skill or local Alembic knowledge base, use the agent-facing prime, work, and Guard tools proactively. For empty projects, do not proactively prime unless the user explicitly asks for Alembic.

Submit new Alembic knowledge candidates from Claude Code when the user explicitly asks, or proactively only when the current project has a project-level Alembic knowledge skill or local Alembic knowledge base. Covers `alembic_submit_knowledge`, V3 fields, batch rules, and review boundaries.

Check code against Alembic Recipe standards with `alembic_code_guard` and explicit files, inline code, or a current workRef with scoped files when the current project has Alembic knowledge.

Discover Alembic project structure, ProjectContext graph orientation, and bounded project-internal relations when the current project has Alembic knowledge. For empty projects, use on explicit Alembic structure requests.

Check code against Alembic Recipe standards with `alembic_code_guard` and explicit files, inline code, or a current workRef with scoped files when this project has local Alembic knowledge.

This project has a local Alembic knowledge base. Submit Alembic candidates when the user asks or when project work reveals reusable patterns, rules, or facts worth preserving.

This project has a local Alembic knowledge base. Use Alembic Recipes, Guard context, and knowledge search proactively for this project when coding or answering project-standard questions.

This project has a local Alembic knowledge base. Use Alembic recipe_map and ProjectContext graph orientation for structure and dependency questions.

Use when working inside ControlWorkspace on AGENTS.md / skill layering, TODO / Backlog intake, Design handoff intake, idle-window scheduling, window coverage, task-package dispatch, producer/consumer sequencing, unified dispatch prompts, test handoffs, validation boundaries, or workspace script pipelines. This skill supplements AGENTS.md and must not override its hard boundaries.

Use when ControlWorkspace total control starts or resumes Codex Automation Closed Loop, reviews target result envelopes, creates dispatch packets, builds delivery envelopes, decides acceptance / rework / block / next wave, or stops unattended automation.

Use when a target Codex window receives a Codex Automation Closed Loop direct-thread delivery, executes only its assigned dispatch packet, reports a TargetResultEnvelope, or enforces target-window boundaries without claim / finish / chain-next state.

Use when ControlWorkspace needs to link into the external Progressive Chain Validation (PCV) canonical source or the local PCVM workspace for source-derived chain plans, scoped rounds, node-level cold-start/rescan baselines, before/after metrics, scorecards, or PCV-guided workflow repair.

Use Alembic Recipes proactively only when the current project has a project-level Alembic knowledge skill or local Alembic knowledge base. For empty projects, use only when the user explicitly asks for Alembic Recipes or setup/status.

Submit new Alembic knowledge candidates from Codex when the user explicitly asks, or proactively only when the current project has a project-level Alembic knowledge skill or local Alembic knowledge base. Covers `alembic_submit_knowledge`, V3 fields, batch rules, and review boundaries.

Use when a Codex control window receives a Feishu/Lark message through Lark Remote, especially prompts containing Lark Remote control messages, remoteCommandId, takeover, observation, or thread dispatch.

Use when the user asks to start, configure, diagnose, or use Lark Remote, or when Codex is responding to a Feishu/Lark message through Lark Remote.

Use when AlembicTest needs to run or monitor real-project Alembic tests, especially BiliDili/AlembicWorkspace cold-start, after-run, Dashboard, PCVM, runtime linkage, or test-mode verification; or when AlembicTest-IDE is explicitly assigned Codex Plugin / host MCP / environment probe evidence. Follow verified routes and collect evidence instead of inventing new paths.

Use when deriving source-based execution plans for long software workflows, validating workflow behavior node by node, repairing current-node failures with explicit evidence and write boundaries, reviewing coverage against a workflow benchmark, or coordinating multi-agent work across a complex chain.