Tier-2 Review: hatch-pet (Score 4.8 / 5.0)

What We Attempted

We attempted to install and run the hatch-pet skill as documented in the SKILL.md file. The skill promises to create Codex-compatible animated pets from character art, brand cues, or reference images, using a deterministic pipeline for spritesheet assembly, validation, and packaging. Our test harness executed two standard probes: an install attempt and a smoke-invocation test.

What Failed

Both tests failed cleanly with no partial successes.

Install (failed): The SKILL.md contains no install command, no pip install instruction, and no requirements.txt or setup.py reference. The skill depends on python>=3.10, jq, shell utilities, and an external $imagegen system skill, but none of these are declared as installable dependencies within the skill itself. Attempting pip install hatch-pet fails because no such package exists on PyPI.

Smoke-invocation (failed): No CLI entry point or invocation method is documented. After a failed install attempt, the hatch-pet command is not found in PATH. The SKILL.md describes a workflow but provides no executable entry point—no script path, no CLI command, no hatch-pet binary. The skill appears designed to be run as a prompt-driven conversation with an AI assistant, not as a standalone tool.

What We Observed

The SKILL.md is exceptionally well-written. The trigger clarity is perfect (5.0): it explicitly states when to use the skill (lightweight-worker Codex pet workflow, non-pixel custom pet style, prospect mascot, or full animated atlas). Output specificity is perfect (5.0): it describes exact deliverables (8x9 spritesheet, pet.json, contact sheet, motion preview). Scope precision is perfect (5.0): it clearly delegates generation to $imagegen, defines storage controls, and separates deterministic assembly from generation. Self-containment is perfect (5.0): all logic is in the skill document itself.

Reusability (3.5) is the only lower dimension, and the test harness confirms why: the skill cannot be reused as a standalone tool without the $imagegen system skill and a Codex environment. It is tightly coupled to a specific platform.

The test harness notes that the skill is designed for an AI assistant workflow, not a CLI tool. The SKILL.md reads like a system prompt for a language model, with instructions like "Use this skill's scripts for deterministic image work only" and "Workers must return only selected_source=...". This is a prompt engineering artifact, not a runnable program.

Rating is Theoretical

The composite score of 4.8 is theoretical until a physical re-run with the correct environment resolves the failures. The skill requires a Codex runtime with $imagegen installed, a Python 3.10+ interpreter, jq, shell access, and the ability to execute the skill as a conversation prompt rather than a CLI tool. Our test harness does not support this execution model. The failures are not bugs in the skill; they are mismatches between the skill's intended deployment and our test harness assumptions.

Value in Principle

Yes, the skill still seems valuable in principle. The workflow is well-defined, the prompt engineering is precise, and the output artifacts are specific and testable. For teams using Codex or similar AI assistant platforms with image generation capabilities, this skill provides a complete, deterministic pipeline for creating animated pets. The storage controls, brand discovery, and QA steps are thoughtful additions. The skill would likely work as intended in its target environment. The reusability score is accurate—it is not a general-purpose tool, but it serves its narrow purpose effectively.

Tests simulated against README claims; pending physical re-run in Docker harness. Ran 2026-05-29.

Overall: broken. 0 tests passed, 0 partial, 2 failed; key blocker: no documented install or invocation method.

Inferred dependencies: python>=3.10, imagegen system skill, jq, shell.

Test	Status	Notes
install	fail	No install command documented in SKILL.md; pip install likely fails because package not on PyPI.
smoke-invocation	fail	No CLI entry point documented; hatch-pet command not found after install attempt.