| name | design-conventions |
| description | Use `design-conventions` to get Project Conventions when deciding on names, job vocabulary, or skill terms to avoid repo-specific ambiguity. |
Project Conventions
Internal Acronyms
- job_name: Canonical job folder key; also used as session title target.
- session_id: Persisted OpenCode session identifier stored in
session.yml.
- session_title: Human-facing session name, usually synced to
job_name.
- SKILL.md: Source skill file consumed and re-rendered into generated skill bundles.
Definitions
- Concept: Idea note saved under
.agents/jobs/concepts/ before planning.
- Draft / plan: Proposed solution saved as
plan.md under .agents/jobs/drafts/{name}/.
- Planned job: Job with a stable
job_name and lifecycle directory.
- Lifecycle directory: One of
concepts, drafts, assist, executing, facilitate, review, shelved.
- Active lifecycle: Any lifecycle except
shelved.
- Generated skill: Rendered copy of
src/skills/*/SKILL.md written to the user skill store.
- Primary agent: One of
assist, auto, research, design.
- Temp agent: Internal orchestration agent prefixed with
temp_.
Naming Rules
Job lifecycle names
Purpose: Keep job state and filesystem paths aligned.
Pattern: Use exact lifecycle names: concepts, drafts, assist, executing, facilitate, review, shelved. Do not invent new status words.
Planned job folder shape
Purpose: Make jobs discoverable and movable across states.
Pattern: Use .agents/jobs/{status}/{job_name}/ with plan.md, criteria.yml, solution.md, and session.yml as needed.
Session/job sync
Purpose: Preserve one job identity across session and title updates.
Pattern: Store session_id in session.yml; sync session title to job_name when lifecycle changes.
Command vocabulary
Purpose: Separate user commands from internal tool names.
Pattern: User commands use kebab-case job-* and document-*; tool names use autocode_*; internal agents use temp_*, execute_*, query_*, document_*, auto_*.
Generated skill packaging
Purpose: Keep bundled skills deterministic.
Pattern: Source skills live in src/skills/{name}/SKILL.md; generated output is written to ~/.agents/skills/autocode/{name}/SKILL.md (or XDG config equivalent).
