mirror of https://github.com/merlinhu1/codex-game-studio.git synced 2026-08-25 07:54:34 +02:00

Files

T

7e8846a7fb feat: add deterministic model tier routing (#14 )

* feat: add deterministic model tier routing

Define Sol, Terra, and Luna policies with fixed fallbacks, explicit surface assignments, runtime precedence, and Luna escalation. Validate generated metadata and expose tier overrides across role, task, and orchestration commands.

* fix: infer model tiers from surface models

* fix: rebalance prompt model routing

* fix: decouple prompt effort routing

---------

Co-authored-by: MerlinH <merlinh221@gmail.com>

2026-07-10 18:49:20 +10:00

2.6 KiB

Raw Blame History

name, description, model, model_reasoning_effort, argument-hint, primary-agent, tool-policy, isolation, source-reference, source-hash, user-invocable

name	description	model	model_reasoning_effort	argument-hint	primary-agent	tool-policy	isolation	source-reference	source-hash	user-invocable
cgs-test-flakiness	Use for test flakiness tasks that diagnose flaky tests by isolating timing, randomness, ordering, environment, and cleanup causes; produce verification evidence, changed or proposed files, and handoff boundaries.	gpt-5.6-terra	high	Describe the test-flakiness objective, target files/assets, constraints, and verification evidence.	qa-playtester	read/edit/shell/tests/git as needed within the repository write policy	repository-root; no init-time generated prompt bodies	.claude/skills/test-flakiness/SKILL.md	3849299d9c0b667a2974157a2ce692892bcef475b87e0c20ad029b22f335b805	true

Codex Game Studio Test Flakiness

Use this skill for test flakiness work in Template Game.

Objective

Diagnose flaky tests by isolating timing, randomness, ordering, environment, and cleanup causes.

Inputs

AGENTS.md
.codex/studio.json
task-relevant files named by the user or task record
design/gdd.md
production/session-state/active.md
tests/

Arguments

Objective or user request.
Target files, scenes, assets, or docs.
Constraints, deadlines, acceptance criteria, and verification command when known.

Procedure

Clarify the requested diagnose flaky tests by isolating timing, randomness, ordering, environment, and cleanup causes. and identify the current project stage.
Collect evidence for Failure Pattern, Isolation, Root Cause.
Run or define the focused validation loop before reporting conclusions.
Produce the requested artifact or review with clear file paths and verification evidence.

Write Targets

tests/
production/session-state/
prototypes/

Output Contract

Summary
Failure Pattern
Isolation
Root Cause
Stabilization
Risks
Changed files or proposed files
Verification evidence
Next owner or decision

Quality Gates

Failure Pattern
Isolation
Root Cause
Stabilization
Scope remains bounded to the current task and project stage.
Report labels unverified assumptions separately from evidence.

Decision Gates

Continue only when the expected output can be verified or clearly labeled as a plan.
Escalate to producer or qa-playtester when scope, ownership, or acceptance evidence is ambiguous.
Stop before broad rewrites, generated prompt mirrors, or hidden lifecycle behavior.

Handoff

Report changed files, verification evidence, remaining risks, and the next owner or decision. Do not imply hidden hooks or autonomous background work.

2.6 KiB Raw Blame History