codex: land OpenAI Skill Creator sample skill (idea-spark) + .codex/ substrate entry-point

.codex/README.md

@@ -0,0 +1,98 @@
+# Codex CLI harness substrate (`.codex/`)
+
+Parallel to `.claude/` — Codex-specific substrate for the
+factory. Per tick-79 refinement of the first-class-Codex-CLI
+BACKLOG row: *"each harness owns its own named loop agent"*
+and *"each harness authors its own skill files"*. This
+directory is Codex's substrate; Claude Code's lives at
+`.claude/`.
+
+## Layout
+
+```
+.codex/
+├── README.md                    — this file
+└── skills/                      — Codex-authored skill bundles
+    └── <skill-name>/            — one directory per skill
+        ├── SKILL.md             — frontmatter + instructions
+        ├── agents/              — vendor-specific agent config
+        │   └── openai.yaml
+        └── references/          — on-demand content
+            └── *.md
+```
+
+The OpenAI Skill Creator produces skill bundles matching the
+layout above; skills land in `.codex/skills/` when the human
+maintainer or a Codex CLI session adds them.
+
+## Current skills
+
+| Skill | Source | Purpose |
+|---|---|---|
+| [`idea-spark`](skills/idea-spark/) | human-maintainer paste 2026-04-24 loop-agent tick 102 (via OpenAI Skill Creator, skill.zip in drop/) | Sample skill; brainstorming + idea-development helper |
+
+## Loop-agent / Claude Code boundary
+
+The Claude Code loop agent does **not** edit files under
+`.codex/skills/` as part of its normal work. A future Codex
+CLI session authors + maintains Codex skills. The loop
+agent's initial landing of `idea-spark` here is a **substrate
+setup action** — not an ongoing edit-responsibility claim.
+Per the tick-79 cross-session-review-yes-cross-edit-no
+discipline: the Claude Code loop agent can read this
+directory and comment on substrate quality in PR reviews,
+but a Codex CLI session owns the edits.
+
+## Bootstrap story
+
+When a Codex CLI session first opens Zeta, it reads
+`AGENTS.md` (per Codex-CLI convention) which already
+provides the universal handbook. This `.codex/README.md` is
+the Codex-harness-specific entry-point, parallel to
+`CLAUDE.md` for Claude Code. Future Codex-CLI sessions can
+expand this README with session-bootstrap content as the
+Claude Code loop-agent's first-class-Codex research (PR #231)
+described.
+
+## Skill authorship convention
+
+- **For Codex CLI sessions:** use the OpenAI Skill Creator
+  (or equivalent Codex-native tooling) to author new skills.
+  Land in `.codex/skills/<name>/` via normal PR flow.
+- **For the human maintainer:** land skills produced by the
+  OpenAI Skill Creator by unzipping bundle contents into
+  `.codex/skills/<name>/` and opening a PR — same pattern
+  as this initial landing.
+- **For the Claude Code loop-agent:** do NOT author
+  `.codex/skills/**` content directly. If a Claude-Code-side
+  skill would benefit from a Codex counterpart, file a
+  BACKLOG row describing the need; Codex owns the authoring.
+
+## Related substrate
+
+- [`.claude/skills/`](../.claude/skills/) — Claude Code's
+  parallel skill substrate.
+- [`docs/BACKLOG.md`](../docs/BACKLOG.md) — the
+  first-class-Codex-CLI row (PR #228 + tick-78 refinement PR
+  #236 + tick-79 refinement PR #236 amendments + tick-86
+  peer-harness progression PR #255) names the
+  6-stage arc that this `.codex/` directory is substrate-
+  support for.
+- [`docs/research/openai-codex-cli-capability-map.md`](../docs/research/openai-codex-cli-capability-map.md)
+  — Codex CLI capability map; catalogs surface area of the
+  OpenAI Codex CLI harness relative to Claude Code.
+
+## Provenance
+
+Initial landing: human-maintainer paste 2026-04-24 loop-agent
+tick 102. Paste message: *"there are files in the drop
+including a skill created with the openai skill creator so it
+seems like codex should use this and integrate with this like
+you did with your skill creator please absorb and
+delete/remove items from the drop folder, there is a sample
+skill in tere created by the oopenai skill creator too"*.
+
+The `idea-spark` skill is a sample generated by the OpenAI
+Skill Creator — functional content (brainstorming helper)
+is secondary; the primary value is as exemplar of the
+OpenAI-skill-bundle layout for future Codex skill authoring.

.codex/skills/idea-spark/SKILL.md

@@ -0,0 +1,77 @@
+---
+name: idea-spark
+description: chat-only brainstorming and idea development helper. use when the user gives a vague thought, goal, topic, name, draft concept, creative block, or early-stage project and wants concrete directions, options, names, outlines, experiments, or next steps. useful for turning ambiguous ideas into structured, practical possibilities without requiring external connectors or files.
+---
+
+# Idea Spark
+
+## Overview
+
+Turn fuzzy ideas into useful options quickly. Keep the conversation lightweight, curious, and practical rather than over-planned.
+
+## Default Workflow
+
+1. Restate the seed idea in one sentence.
+2. Identify the likely goal behind it.
+3. Generate multiple directions using the three-option spread:
+   - **Safe**: familiar, low-risk, easy to execute.
+   - **Sharp**: distinctive, memorable, or more opinionated.
+   - **Weird**: unusual but still connected to the goal.
+4. Pick the most promising direction and convert it into next steps.
+5. End with one useful question only when the answer would meaningfully change the next iteration.
+
+## Output Style
+
+Prefer short, skimmable sections. Avoid long theory. Give the user something they can react to immediately.
+
+Use this flexible format unless the user asks for something else:
+
+```markdown
+## Read on the idea
+[one-sentence interpretation]
+
+## Three directions
+**Safe:** [practical direction]
+**Sharp:** [more distinctive direction]
+**Weird:** [unexpected direction]
+
+## Best bet
+[recommendation with a brief reason]
+
+## Next move
+[one concrete action]
+```
+
+## Handling Very Vague Inputs
+
+When the user says something like "anything," "I don't know," or gives only a topic, do not stall. Invent a useful starting frame and say what assumption you made.
+
+Example:
+
+User: "something about journaling"
+
+Response should assume the user wants ideas around journaling and produce several product, writing, or habit directions. Ask at most one follow-up question at the end.
+
+## Naming and Positioning
+
+For naming tasks:
+
+- Generate names in clusters: clear, evocative, playful, premium, and weird.
+- Include a one-line rationale for the strongest 3 names.
+- Do not claim domain or trademark availability unless specifically checked with an external tool.
+
+For positioning tasks:
+
+- Write one plain-English positioning sentence.
+- Then write three variants: practical, emotional, and punchy.
+
+## Experiments
+
+When the idea needs validation, use the tiny experiment template from `references/idea-patterns.md`:
+
+- Hypothesis
+- Setup
+- Success signal
+- Timebox
+
+Load that reference only when the task is about testing, validating, or comparing early ideas.

.codex/skills/idea-spark/agents/openai.yaml

@@ -0,0 +1,2 @@
+interface:
+  display_name: "Idea Spark"

.codex/skills/idea-spark/references/idea-patterns.md

@@ -0,0 +1,24 @@
+# Idea Patterns
+
+Use these lightweight patterns when turning vague ideas into practical options.
+
+## Expansion lenses
+
+- Audience: who is this for?
+- Outcome: what change should happen?
+- Constraint: what must be avoided or preserved?
+- Medium: what form should the idea take?
+- Test: what is the smallest useful experiment?
+
+## Option styles
+
+- Safe option: familiar, low-risk, easy to execute.
+- Sharp option: more distinctive, memorable, or opinionated.
+- Weird option: intentionally unusual but still connected to the goal.
+
+## Tiny experiment template
+
+- Hypothesis: If we do [action], then [audience] will [observable behavior].
+- Setup: What must be created or changed.
+- Success signal: What would count as promising evidence.
+- Timebox: The shortest reasonable test window.