DexForce/EmbodiChain / .agents/skills/add-task-program
add-task-program
VerifiedCreate, 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.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 5.4 KB |
| agents/openai.yaml | config | 102 B |
| references/deployment.md | markdown | 3.3 KB |
| references/integration-composition.md | markdown | 4.4 KB |
| references/program-authoring.md | markdown | 3.0 KB |
| references/validation.md | markdown | 2.9 KB |
| scripts/inspect_deployment.py | script | 22.6 KB |
references/program-authoring.md
View on GitHub# Program Authoring
Author task intent independently of robots, simulators, and live providers.
## Choose serialized or Python form
Use YAML/JSON when the program is a deployable task asset or should be authored
by an agent/MLLM. Use Python `TaskProgramCfg` when the caller explicitly wants
programmatic composition, reuse in code, or a typed tutorial/test.
Serialized programs belong at:
```text
extensions/embodichain_tasks/configs/tasks/<category_path>/<task_name>/task_program/program.yaml
```
The public Python API is `embodichain.lab.task_program`. A provider-independent
Python example is:
```text
scripts/tutorials/task_program/build_and_compile.py
```
## Program contract
A program owns:
- one stable `program_id`;
- optional named targets;
- one bounded program tree.
Supported node families are sequence, repeat, segment, invoke, and parallel
with an owned barrier. Built-in call configs are Pick, Place, HandOver, and MoveHeldObject. List/action
shorthand expands to the same strict AST; objects stay explicit and string
`at`/`final_target` fields select named targets. Targets support world poses,
scene-relative poses and repeat-local cyclic poses. Typed transport and duration
`hold` policies require verified effects. Use `demo_skill: {skill, roles}` only
with an alias declared by the selected integration's `runtime_services.demo_skills`;
composition generates its content-pinned occurrence sources.
`RegisteredSemanticCallCfg` is the extension payload for an allowlisted call.
Do not serialize the trusted integration selection. It is injected from the
runnable deployment after component composition.
## Authoring sequence
1. Express the minimum ordered task outcome as Semantic Calls.
2. Add explicit segments where post-policies, validators, instructions, or
dataset boundaries matter.
3. Add named/cyclic targets only when multiple calls reuse or advance them.
4. Use bounded repeats instead of generated duplicate steps.
5. Add parallel branches only when resource claims, runtime targets, control
grids, symbolic writes, and physical safety can be proven compatible.
6. Add validators for dataset/task acceptance; do not confuse them with
per-call effect monitors.
Use exact logical scene IDs from the planned integration, not native simulator
UIDs unless the mapping intentionally uses the same string.
## Safety and limits
The strict decoder rejects unknown fields, duplicate keys, non-finite values,
invalid exact types, unresolved references, excessive depth/nodes/repeats, and
executable registered payloads. Do not bypass it by constructing mutable
internal objects.
Nested parallel blocks are unsupported. Parallel resource disjointness alone
is not physical collision-safety evidence.
## Validation
For program-only work, run focused language/compiler tests and load the file
with `load_task_program()`. Full deployment validation additionally needs the
trusted integration selection and catalog; use the deployment inspector after
the other components exist.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deployment | Stars: 319k | Updated | Claude CodeCodex | Verified | Install |
| 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: 196k | Updated | Claude CodeCodex | Verified | Install |
| 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: 118k | Updated | Claude CodeCodex | Verified | Install |
| 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: 117k | Updated | Claude CodeCodex | Verified | Install |