Menu
Convert a PyAutoMind prompt file into a GitHub issue and register it — the Mind issue and registry primitive that start_dev delegates to.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
| name | create-issue |
| description | Convert a PyAutoMind prompt file into a GitHub issue and register it — the Mind issue and registry primitive that start_dev delegates to. |
Turn a PyAutoMind/ prompt into a tracked GitHub issue and register it.
A PyAutoMind primitive — it owns the mechanical issue + registry write:
assemble the issue body, create the issue, register the task in active.md, move
the prompt to active/, and push Mind state. The reasoning (classify repos,
explore code, generate the plan) belongs to PyAutoBrain:
$start-dev (/start_dev in Claude) does the triage/planning via the
Feature Agent and then delegates the issue write to this skill, passing the
primary repo, title, plan and suggested branch. Brain does not re-implement
issue creation.$create-issue (/create_issue in Claude) by itself when you just
want an issue from a prompt without full triage or worktree routing — it does a
light pass to fill any inputs the caller didn't supply. For deep
classification + dev-environment setup, use $start-dev instead.Organ boundary and the execution-environment model: PyAutoBrain skills/WORKFLOW.md.
$create-issue <prompt-file-path> # /create_issue in Claude
Path relative to PyAutoMind/. Prompts live under <work-type>/<target>/ (see
README "Prompt taxonomy"); pre-migration <target>/<name>.md paths still resolve.
Examples: bug/autofit/factor_graph_instance_iteration.md,
feature/autoarray/oversampling.md.
These come from the caller (Brain/$start-dev) when delegated, or from a
light standalone pass here when run directly:
| Input | From caller | Standalone fallback |
|---|---|---|
| primary repo | Feature Agent classification | most-referenced @RepoName in the prompt; ask if ambiguous |
| title | caller | concise title (<70 chars), conventional prefix (feat:/fix:/refactor:/docs:/perf:) |
| plan (high + detailed) | caller | brief plan from a quick read of the prompt + referenced files |
| suggested branch | caller (plan_branches) | feature/<short-desc> kebab-case, <50 chars |
Resolve repository owners from PyAutoMind/repos.yaml; the generated summary
in WORKFLOW.md is the readable mapping. Do not reuse legacy owner defaults.
source PyAutoMind/scripts/prompt_sync.sh
prompt_sync_new_prompts # no-op if nothing untracked; else commits + pushes new ideas
Read PyAutoMind/<argument>. If missing, report and list prompts in that folder.
If the caller supplied repo/title/plan/branch, use them verbatim. Otherwise
do the light standalone pass from the Inputs table (don't run a full Brain
triage — that's $start-dev).
Build the body in this structure, then create it (present for review first):
## Overview
<2-4 sentence summary of what this task is and why it matters>
## Plan
<high-level bullet plan — human readable, no code>
<details>
<summary>Detailed implementation plan</summary>
### Affected Repositories
- repo1 (primary)
### Branch Survey
| Repository | Current Branch | Dirty? |
|-----------|---------------|--------|
| ./RepoName | main | clean |
**Suggested branch:** `feature/<name>`
### Implementation Steps
1. <step with file paths>
### Key Files
- `path/to/file.py` — description
</details>
## Original Prompt
<details>
<summary>Click to expand starting prompt</summary>
<original prompt content copied verbatim>
</details>
gh issue create --repo <owner/repo> --title "<title>" --body "$(cat <<'ISSUE_EOF'
<body content>
ISSUE_EOF
)"
Add the task entry to PyAutoMind/active.md with the issue URL (schema in
README). If the caller is handling registration itself — e.g. $start-dev
routing a conflicted task to planned.md — skip this step and let it register.
The prompt advances from draft/ (not started) to active/ (issued, in
flight) — the second of the three lifecycle states (draft/ → active/ → complete/; issue #71). Use git mv to preserve history:
mkdir -p PyAutoMind/active
git -C PyAutoMind mv <draft-path> active/<filename>
The mkdir -p matters on a freshly-spawned Mind, where active/ does not
yet exist (it holds only instance state, so the template ships without it) —
git mv errors if the destination dir is missing. Timestamp-suffix the
filename if one already exists in active/.
source PyAutoMind/scripts/prompt_sync.sh
prompt_sync_push "prompt: file issue for <task-name> (#<issue>)"
If step 0 already pushed, this carries only the active.md + active/ changes.
gh auth status fails, tell the user to run ! gh auth login.