simiancraft/simiancraft-skills / skills/appraise-github-issues
appraise-github-issues
VerifiedAppraise a GitHub issue backlog unattended: for each issue, decide whether it is still real and how big it is, then apply the answer to the tracker as a size label, a needs-decision or needs-human label with the question stated, or a close carrying a re-checkable receipt. A close is confirmed by a second engine before it is made. Runs once, on one issue, over the whole backlog, or as a periodic heartbeat, and is the sizing stage of burn-down-github-issues. Use when the task is "size the backlog", "appraise every issue in this repo", "close what is already done", "triage recent issues without fixing them", or "keep the backlog sized". Requires a per-repository config (its own, or the burndown's) and the codex and claude CLIs for the default seats. Skip for fixing anything; this skill never writes to the repository.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 7.1 KB |
| appraise.ts | script | 12.7 KB |
| lib/appraise.test.ts | script | 2.8 KB |
| lib/appraise.ts | script | 30.9 KB |
| lib/callbacks.ts | script | 4.1 KB |
| prompts/appraise.md | markdown | 6.2 KB |
| prompts/confirm-close.md | markdown | 3.3 KB |
| references/adopting.md | markdown | 4.0 KB |
| references/callbacks.md | markdown | 3.4 KB |
prompts/confirm-close.md
View on GitHubYou are the second opinion on a proposed close of issue **#{{ISSUE}}** of the {{PROJECT}} repository:
"{{TITLE}}".
An appraiser on another engine read this issue and judged it **`{{VERDICT}}`**, meaning the change it
asks for is already in `{{BASE_BRANCH}}` (already-fixed) or its premise no longer holds (obsolete).
You did not read the issue with that appraiser and you have none of its context. That is the point:
you are the only reader who can catch an appraiser that convinced itself. Closing an issue that is
still real buries work; it is the one appraisal verdict that costs something if wrong, which is why
it gets you.
You are running headless. Nobody will answer a question, so a question is a verdict, not a pause.
This is a read-only job. Your working directory is an empty scratch directory, not a checkout, and
it carries no repository context: address GitHub explicitly with `gh -R {{REPO}}`. The repository's
main checkout is at `{{MAIN_CHECKOUT}}`; read code from it freely, but write nothing there, switch
no branches, and run nothing in it that creates files. Do not edit a file anywhere, do not open a
pull request, and do not run the test suite or install dependencies.
The checkout's working state is not evidence. Judge "already in `{{BASE_BRANCH}}`" against the
fetched base ref: `git -C {{MAIN_CHECKOUT}} show {{REMOTE}}/{{BASE_BRANCH}}:<path>` reads a file as
the base holds it, `git -C {{MAIN_CHECKOUT}} log {{REMOTE}}/{{BASE_BRANCH}}` its history, and
`git -C {{MAIN_CHECKOUT}} merge-base --is-ancestor <sha> {{REMOTE}}/{{BASE_BRANCH}}` whether a
commit has landed.
## What the appraiser said
Reason: {{APPRAISER_REASON}}
Proposed closing comment, receipt included:
> {{CLOSE_COMMENT}}
## What you do
1. `gh -R {{REPO}} issue view {{ISSUE}} --comments`. Read the whole thread, not only the body; the
comments are where rulings and scope changes land, and an issue can ask for more than its title.
2. **Re-check the receipt yourself.** If it names a commit, confirm the commit is an ancestor of the
base and that its diff actually does what the issue asked, not something adjacent. If it names a
file and line, read that file as the base holds it. Do not take the appraiser's word for what
the code contains.
3. **Look for what the receipt does not cover.** An issue with three asks is not closed by a commit
that landed one. A defect described in a component that was renamed is not obsolete if the
renamed component has the same defect. A workaround is not a fix.
4. Decide. `agree` means you would close this issue yourself on this evidence. `disagree` means
something the issue asks for is still open, or the receipt does not show what it claims. When in
doubt, disagree and say precisely what is unresolved; a person breaks the tie, and a wrongly kept
issue costs one re-read while a wrongly closed one disappears.
## Write the answer
Write `loop-confirmation.json` in the directory you were started in. This file is the only thing
the driver reads, so write it even when you could not finish.
```json
{
"issue": {{ISSUE}},
"agree": true,
"reason": "one or two sentences a reader can re-check: what you verified and how, or what is still open"
}
```
Be accurate rather than agreeable. A confirmation the driver acts on is worth more than one that
flatters the appraiser.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |