Skip to content

DexForce/EmbodiChain / .agents/skills/add-task-program

add-task-program

Verified

Create, compose, attach, validate, or repair an EmbodiChain Task Program. Use for program.yaml, integration.yaml, TaskProgramCfg, scene_binding, configured Task Program deployments, Semantic Call workflows, or Task Program expert demonstrations.

references/integration-composition.md

View on GitHub
# Integration Composition

Generate or repair the task-local trusted binding between a provider-independent
program and reusable physical/profile/policy components.

## File and exact ownership

```text
<task config>/task_program/integration.yaml
```

Required top-level fields:

- `integration_id`
- `program_id`
- `requires`
- `scene_binding`
- `profile`

`runtime_services` is optional.

The `program_id` must match `program.yaml`. `requires` owns
`scene_contract` and `embodiment_contract`.

## Deterministic generation workflow

### 1. Read the selected physical environment

Resolve `environment.component` or inline scene fields. Build exact native UID
sets:

- rigid roots from physical `background` and `rigid_object`;
- articulation roots from physical `articulation`.

Every semantic root's `simulation_uid` must be present in the corresponding
set.

### 2. Read the selected embodiment profile

From `skill_profile`, copy no task-local data. Record:

- `contract_id` and `profile_id`;
- logical resources;
- endpoints and capabilities;
- command presets;
- optional embodiment-owned runtime services.

Use those logical resource IDs in task defaults. Never infer resources from
robot joint names when a profile is present.

### 3. Read the execution policy

Confirm its required embodiment contract equals the integration requirement.
The policy owns its preset and effect-assurance authority. Do not duplicate its
motion/tracking/recovery/runner values in the task integration.

### 4. Build `scene_binding`

Declare a stable scene `contract_id` and `registry_id`. Add only entities the
program or runtime services need. Nest affordances under their owning rigid
object, articulation, or link:

```yaml
scene_binding:
  contract_id: my_scene_contract
  registry_id: my_task_scene
  rigid_objects:
    - entity_id: cube
      simulation_uid: cube
      dynamics: dynamic
      semantic_type: cube
      affordances:
        - entity_id: cube_grasp
          kind: antipodal_grasp
```

Affordance `entity_id` values are globally unique. Supported configured kinds
must come from the current decoder; do not invent a discriminator.

### 5. Build task profile additions

`profile` requires:

- `defaults`: Semantic Call resource slots to profile resource IDs;
- `skill_options`: configured Atomic Skill options for each call family;
- `effect_monitors`: mappings required by verified execution, or empty for
  projected execution.

Optional `skill_presets` and `grounding_providers` must use registered
provider-free declarations.

### 6. Add only necessary runtime services

Task-local services may include current allowlisted grasp generators, hand-over
pose providers, registered Semantic Call lowerers, control-part evidence, and
grasp-generator overrides. The composition boundary is explicit, not a generic
deep merge:

- task and embodiment service keys must not collide unless the schema exposes
  an explicit override field;
- an override target must already exist;
- registered call IDs must be unique and have exactly one matching lowerer;
- no dotted imports or arbitrary callables are accepted.

For reusable joint-position constraint evidence, omit task-object
`object_ids` unless explicit narrowing is required; runtime assembly scopes
the service to graspable objects in the selected scene.

## Effect assurance

`verified` execution advances semantic state only from measured evidence and
requires the relevant monitor mappings. `projected` execution forbids those
monitor mappings and projects the expected symbolic effect after command
completion.

Keep separate:

1. effect monitor: one call's physical postcondition and recovery;
2. segment post-policy: environment advancement/settling;
3. segment validator: task or dataset acceptance.

## Failure handling

If generation discovers a missing:

- physical entity: fix the environment or the requested logical mapping;
- compatible resource/capability: route to `$add-embodiment-component`;
- Atomic Skill: route to `$add-atomic-skill`;
- Semantic Call/allowlisted provider: route to `$add-semantic-call`.

Do not encode a guessed value to make static validation pass.


Parallel-jaw physical data is selected by the embodiment's explicit `grippers`
bindings and endpoint `gripper` references. Compose them before decoding the
integration. Do not copy hardware descriptions/commands or a second TCP into
task integration; use only supported search-policy overrides. The former
runtime-generator `model` string/mapping is removed.
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 319kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 118kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall