Neural Codex is one repository marketplace containing one plugin. Its supported
surface is exactly five skills: discover, spec, craft, vet, and
exercise, plus the reviewed hooks under plugins/neural-codex/hooks/.
Do not add a second skill root, a second hook manifest, or an installer that copies files into a user's home directory. The plugin and marketplace are the distribution contract.
- Marketplace:
.agents/plugins/marketplace.json - Plugin manifest:
plugins/neural-codex/.codex-plugin/plugin.json - Skills:
plugins/neural-codex/skills/<name>/SKILL.md - Hooks:
plugins/neural-codex/hooks/hooks.json - Public docs:
docs/ - Plans and evidence:
plans/
Use $discover for material ambiguity, $spec before non-trivial
implementation, $craft for an approved plan, and independent $vet plus
$exercise gates before shipping. Keep generated evidence beside its plan.
Never auto-trust hooks or mutate a user's global Codex configuration. Never
commit credentials or read/write .env contents as part of validation.
python3 -m pytest -q
./scripts/doc-lint.sh
python3 /path/to/plugin-creator/scripts/validate_plugin.py plugins/neural-codexValidate each retained skill with Codex's quick_validate.py when that tool is
available. A valid directory shape is necessary but not sufficient: tests must
also reject stale claims and unsupported inventory.
- Architecture:
ARCHITECTURE.md - Harness contract:
docs/AGENT-HARNESS.md - Workflow:
docs/WORKFLOW.md - Hooks and trust:
docs/HOOKS.md - Model availability and prompting:
docs/CONFIGURATION.md - Verification:
docs/VERIFICATION.md