haipipe-task-for-algo

name	haipipe-task-for-algo
description	algo-dev task-folder specialist: scaffolds {NN}_<name>/ task-folders in the algo-dev task-group (default X_algo) that smoke-test a newly developed algorithm class end-to-end on a TINY config. NOT full training -- see /haipipe-task-for-fit. Called by /haipipe-task when task-type=algo. Cross-references /haipipe-nn-algo.
argument-hint	[project_id] [group] [task-name]
allowed-tools	Bash, Read, Write, Edit, Grep, Glob, Skill
metadata	{"version":"0.1.2","last_updated":"2026-07-04","summary":"algo-dev task-folder build specialist."}

Skill: haipipe-task-for-algo

Scaffolds an algo-dev smoke-test task-folder. Purpose: verify a new algorithm class (forward / loss / metric) runs end-to-end. This is NOT a training run — minimal config, minutes-not-hours, just "did it crash + does the loss go down on one batch".

Invocation modes: interactive (human steers; missing fields get ASKed) OR headless (haipipe-task-creator-agent calls this skill during Phase 2: Build, then authors the <TASK>.py body). Always end with the structured return block (status / task_folder / run_name / files).

task-algo vs task-training (don't confuse them)

                    task-algo (this)        task-training
Purpose             smoke-test algorithm    train + sweep a model
Group letter        X (paired demo)         A (model-run)
Config              minimal / 1-batch       full hyperparam grid
Runtime             minutes                 hours-to-days
Outputs             "didn't crash" + loss   checkpoint → _WorkSpace/5
Audience            algo developer self     cross-run comparison
Pipeline skill      /haipipe-nn-algo        /haipipe-nn-tuner+instance

What this scaffolds

tasks/X_algo/                                ← X-series group (paired Track A)
└── {NN}_test_<algo_name>/
    ├── {NN}_test_<algo_name>.py
    ├── configs/
    │   └── algo_<name>_tiny.yaml            seeded from ref/config-seed.yaml
    ├── runs/
    │   └── algo_<name>_tiny.sh
    ├── results/                             loss.json, "ran" marker
    └── notebooks/

Group letter default: X (algo-dev demo). Heavy outputs: none (tiny / disposable).

Cross-reference to pipeline skill

/haipipe-nn-algo owns the algorithm class itself (Layer 1: model, forward, loss, metric). This skill scaffolds the smoke-test demo that exercises it. Typical flow:

1. /haipipe-nn-algo — author the algorithm class.
2. /haipipe-task-for-algo — scaffold the paired demo.
3. Run the demo. Iterate.
4. Once stable → /haipipe-task-for-fit for the real run.

Scaffold flow

See fn/scaffold.md for the detailed step-by-step. Summary:

1. Identify project + task-group.
2. Collect metadata (NN, name, type-specific extras, _meta block).
3. Create skeleton (.py, configs/, runs/, results/, notebooks/).
4. Seed config from ref/config-seed.yaml.
5. Copy run-script from ../../haipipe-task/ref/run-sh-template.sh.
6. Suggest next via cross-skill link.
7. Emit return contract.

Return contract

status:    ok | blocked | failed
summary:   2-3 sentences on what was scaffolded
artifacts: [paths created]
next:      suggested next command (run the demo / /haipipe-nn-algo refine)

Workflow plan

When /haipipe-task plan targets an existing task-folder of this type, the generated plan-script YAML should follow the type-specific sample:

ref/workflow-plan-sample.yaml     ← script-level phases for this type
../../haipipe-task/ref/workflow-template.yaml  ← task-level template (Run/Gate1/Gate2)

Schema source of truth: task/haipipe-workflow/ref/plan-schema.md