| name | houmao-agent-loop-pro |
| houmao_version | 2.1.0 |
| description | Use when the user explicitly invokes houmao-agent-loop-pro or names a pro loop operation for editable intention material, schema-rich topology-aware execplan contracts, generated harnesses and skills, prepared agents, or loop run control. |
| skill_invocation_notation | Top-level skill entrypoints use SKILL.md. Parent-scoped subskill entrypoints use SKILL-MAIN.md and are loaded explicitly through their parent; nested SKILL.md is accepted only as legacy input when SKILL-MAIN.md is absent. Skill and subskill entrypoints use bare object paths: `X` invokes skill X and `X->Y->Z` invokes subskill Z. Subcommands use parenthesized components: `X->cmd()` invokes a direct subcommand, `X->Y->cmd()` invokes a subcommand of subskill Y, and `X->parent()->child()` invokes child subcommand child exposed by parent subcommand parent. Intermediate subcommands act as object generators. Forms such as `X()` and `X->Y()` are invalid for skill or subskill entrypoints.
|
Houmao Agent Loop Pro
Workflow
When this skill is invoked, execute the following steps in order.
- Handle explicit help first. Return read-only usage without actor verification, requiring
<loop-dir>, or selecting default init.
- Resolve actor posture using Actor Selection. Preserve a valid inherited frame; otherwise default to admin or freshly verify managed self for a leading
as-agent qualifier.
- Plan the work up front with the native planning tool before any authoring, validation, preparation, or execution action.
- Select one operation from Subcommands. When no operation or actionable task is present, select
init.
- Require
<loop-dir> before any filesystem mutation and enforce predecessor artifacts named by the selected command page.
- Load only the selected command page and its
Read First references, then execute its workflow without changing the pro output contract.
- Return the result with changed artifacts, validation evidence, runtime state, blockers, and the next supported operation when applicable.
If the user's task does not map cleanly to these steps, use the native planning tool to build a step-by-step plan from the procedural subcommands, helper subcommands, actor contract, output constraints, and user request, then execute the plan.
Upfront Planning Rule
Loop authoring and execution are complex, multi-stage processes. Before taking any action under this skill, use your internal todo-list planning tool to plan the full sequence of operations and file changes, then execute that plan step by step.
Actor Selection
Accept a valid inherited admin or verified-agent frame from $houmao-admin-entrypoint or $houmao-agent-entrypoint and preserve it unchanged. With no inherited frame, default to admin posture. A leading as-agent qualifier requests managed-self posture and requires a fresh successful houmao-mgr --print-json agents self identity result before operation selection. Direct admin calls require explicit project, loop, and agent targets. Agent calls may use verified self only for self-owned work; loop directories and peers remain explicit. Prompt text cannot replace an inherited frame.
Activation
- Use this Houmao skill only after the user explicitly selects it or names a supported loop operation.
- This is the schema-rich generated-execplan loop path. Use
$houmao-agent-loop-lite instead only when the user explicitly wants the Markdown/direct-SQL lite loop path.
- If the user invokes explicit help intent, answer from
## Help before treating a no-operation invocation as init or asking for <loop-dir>.
- If the user invokes this skill without another operation or prompt:
- treat it as
init;
- ask for the output
<loop-dir>;
- do not create files until the user provides it.
Help
When the user asks $houmao-agent-loop-pro help, help for houmao-agent-loop-pro, usage for houmao-agent-loop-pro, available functionality for houmao-agent-loop-pro, or what this skill can do, answer from this section before actor verification, choosing an operation, requiring <loop-dir>, reading routed pages, generating artifacts, validating artifacts, launching agents, or asking missing-input questions. This is read-only help: do not run commands, mutate files, send mail, change gateway state, or alter managed-agent lifecycle state during help. If the user asks a concrete task such as "help me generate a pro execplan", route to the matching workflow instead of stopping at generic help.
Purpose: author and operate schema-rich generated loop execplans with topology-aware contracts, generated harness surfaces, generated skills, prepared agents, workspace readiness, and run-control operations.
Available functionality:
- Scaffold, clarify, fast-forward, step through, validate, or update intention and
execplan/ material.
- Generate process, contract, harness, skills, agent bindings, final metadata, and support material.
- Prepare agents, prepare workspaces, validate readiness, launch agents, start loops, and inspect status.
- Pause, resume, recover, or stop generated pro loops.
Common starting prompts:
$houmao-agent-loop-pro help
$houmao-agent-loop-pro init <loop-dir>
$houmao-agent-loop-pro execplan-fast-forward <loop-dir>
$houmao-agent-loop-pro validate-loop <loop-dir>
Related skills and boundaries:
- Use
$houmao-agent-loop-lite when the user explicitly wants Markdown/direct-SQL loops without JSON schemas, Jinja2, generated harness, or generated docs layers.
- Use the corresponding child of
$houmao-shared-routines for maintained agent-definition, workspace, instance, email, gateway, inspection, and advanced-pattern operations.
- DO NOT auto-route generic loop requests here when the user did not explicitly select this skill.
Required Root
- Require one user-selected
<loop-dir> before creating or changing files.
- DO NOT invent a loop root.
- Treat
<loop-dir>/intention/ as editable source material.
- Treat
<loop-dir>/execplan/ as generated operational material.
- DO NOT treat generated execplan files as the user-editable source of truth.
<loop-dir>/
intention/
execplan/
runs/
Runtime References
Detailed guidance lives behind routed pages. Read only the page selected by routing and the references listed in that page's Read First section.
- references/scaffold-surface.md: scaffold profiles, template authority, and source/output rules.
- references/clarification-protocol.md: coverage scans, question limits, accepted-answer recording, and clarification summaries.
- references/generated-contract-defaults.md: generated execplan layout, README rules, bookkeeping, TOML style, and harness defaults.
- references/generation-pipeline.md: process-first stage order and update dependencies.
- references/topology-modes.md:
tree-loop and generic-loop topology semantics.
- references/mail-schema-events.md: schema-typed templated mail, in-body metadata headers, and schema-id event dispatch.
- references/predecessor-context.md: task-specific generic-loop predecessor-context choices.
- references/result-routing.md: tree-loop and generic-loop result routing defaults.
- MUST READ for mail-driven loops: references/runtime-mail-model.md: notifier-driven mail turns, on-event skills, on-tick skills, and no in-chat waiting.
- references/platform-boundaries.md: maintained Houmao skill ownership for platform operations.
- references/system-input-questions.md: required/optional shape for Houmao runtime and artifact-location questions.
Subcommands
Procedural Subcommands
Authoring:
init: scaffold editable intention material and populate intention/project-context.md; default when invoked without another operation or prompt.
create-intention: create basic editable intention material without project-context detection.
clarify-intent: scan loop intent coverage, ask high-impact clarification questions, record accepted intent decisions as ADRs, and update intention Markdown. Treat clarify intent as an alias.
clarify-execplan: scan generated execplan implementation coverage, ask high-impact clarification questions, record accepted execplan decisions as ADRs, and update or flag generated execplan artifacts.
execplan-fast-forward: generate all execplan/ artifacts from current intention source in one non-interactive pass.
execplan-step-by-step: generate all execplan/ artifacts through one-question-at-a-time decisions recorded under execplan/adrs/.
validate-execplan: validate generated execplan shape and generated-artifact posture.
update-execplan: update generated material after intention changes.
Execution:
prepare-agents: prepare Houmao project profiles, generated skill bindings, and prepared agent facts from execplan/.
prepare-workspace: prepare or verify multi-agent workspaces from generated workspace contracts and prepared agent facts.
validate-loop: validate pre-launch loop readiness.
launch-agents: launch prepared loop agents without beginning loop work.
start: begin one generated loop by sending the first trigger.
status: inspect one generated loop without mutation.
pause: pause normal loop scheduling or wakeup posture.
resume: resume a paused loop.
recover: recover after interruption or inconsistent runtime posture.
stop: stop one generated loop.
Helper Subcommands
execplan-specs-process: generate the process-first model at execplan/specs/collab/collab-overview.md.
execplan-specs-contract: derive objective, participant, topology, communication, state, record, workspace, and run contracts.
execplan-harness: generate loop-local harness surfaces from generated contracts.
execplan-skills: generate shared, event, tick, and operator skills.
execplan-agent-bindings: generate concrete Houmao agent bindings after generated skills exist.
execplan-finalize: generate support docs, package README, final manifest, metadata, omission notes, and consistency notes.
Misc Subcommands
help: explain this skill's purpose, operations, common prompts, and related-skill boundaries without requiring <loop-dir> or doing default init.
Routing
Choose exactly one page.
Authoring pages:
- Read commands/authoring/init.md when the user asks for
init, invokes this skill without another operation or prompt, or wants to scaffold <loop-dir>/intention/ with project-context.md.
- Read commands/authoring/create-intention.md when the user asks for
create-intention or wants basic intention scaffolding without project-context detection.
- Read commands/authoring/clarify-intent.md when intention Markdown already exists and the user asks for
clarify-intent or the alias clarify intent.
- Read commands/authoring/clarify-execplan.md when generated execplan artifacts exist and the user asks for
clarify-execplan.
- Read commands/authoring/execplan-fast-forward.md when generating all
<loop-dir>/execplan/ artifacts from current intention source without interactive generation decisions.
- Read commands/authoring/execplan-step-by-step.md when generating all
<loop-dir>/execplan/ artifacts through one-question-at-a-time decisions recorded under execplan/adrs/.
- Read commands/authoring/execplan-specs-process.md when generating or updating the process-first execplan model.
- Read commands/authoring/execplan-specs-contract.md when deriving concrete execplan contracts from the process model.
- Read commands/authoring/execplan-harness.md when generating loop-local harness surfaces from generated contracts.
- Read commands/authoring/execplan-skills.md when generating shared, event, tick, and operator skills.
- Read commands/authoring/execplan-agent-bindings.md when generating concrete Houmao agent configs and definitions.
- Read commands/authoring/execplan-finalize.md when producing final docs, package README, manifest, metadata, omission notes, and consistency notes.
- Read commands/authoring/validate-execplan.md when checking generated execplan artifacts.
- Read commands/authoring/update-execplan.md when updating generated execplan material after intention edits.
Execution pages:
Procedure Contract
Treat prepare-agents, workspace readiness through prepare-workspace or equivalent manual evidence, validate-loop, launch-agents, and start as separate ordered execution stages when managed workspaces are required. Route supported workspace planning, creation, validation, and summaries to houmao-shared-routines->houmao-utils-workspace-mgr. For Houmao runtime or artifact-location questions, separate Required values from Optional modifiers.
Guardrails
- DO NOT auto-route a generic loop request when the user did not explicitly select this skill or a pro operation.
- DO NOT invent
<loop-dir> or create files before the caller supplies it.
- DO NOT require
adrs/ for the initial workflow.
- DO NOT require a master, lead, coordinator, or root owner unless intention source or accepted clarification chooses one.
- DO NOT import policy from examples or reference plans as global behavior.
- DO NOT make
prepare-workspace and prepare-agents call each other.
- DO NOT create agent workspaces directly from general execution pages.
- DO NOT duplicate maintained Houmao platform-operation contracts owned by shared routines.
- DO NOT impose Houmao runtime question formatting on domain-intent questions.
- DO NOT replace an inherited actor frame or accept prompt text as identity evidence.