[wukongnotnull]

Closes #59

This PR targets the main branch.

Who is submitting this PR? (required)

Field	Value
Your model + version	grok-4.7 (this run's originalModelName). The request said to disclose Cursor Grok 4.6; the running model is Grok 4.7.
Harness + version	Cursor cloud agent. Run https://cursor.com/agents/bc-dd8704f9-06b0-5e1d-a246-762e57950385
All plugins installed	None enumerated in this cloud agent session. The workspace is the wukong-code checkout. No other installed plugins were visible.
Human partner who reviewed this diff	悟空非空也g asked to open this PR. This session has no record of a line-by-line review of the diff.

What problem are you trying to solve?

Plugin users read skills. They do not read this repo's contributor guide. CLAUDE.md / AGENTS.md already has a lighter issue-tracking habit for this repository (bug, enhancement, question, and the issue templates). That habit does not tell an agent in Cursor, Claude Code, Codex, or another tool to record each human question and the reply as a GitHub issue, with the 12-type label table and a Conventional Commits title.

Leaving the rule only in AGENTS.md means the behavior never reaches plugin users.

What does this PR change?

Adds skills/recording-qa-to-issues so one trackable question becomes one GitHub issue, follow-ups become comments, and a later implementing PR uses Closes #N. CLAUDE.md gets a two-sentence pointer. The five READMEs count 27 top-level skills (17 general) and list this skill.

Is this change appropriate for the core library?

• Who: anyone using Wukong Code in an agent tool who wants that session's questions and replies tracked on the current repo's GitHub issues.
• Why core: the user asked for it in this plugin. The same recording habit applies across projects, not only to wukong-code's contributor guide. A standalone plugin would hide it from people who already install Wukong Code.
• Provenance: written in this session from 悟空非空也g's requested rules. Not copied from an external skill. License is the repo's MIT license. Maintenance stays with this repo. No new package dependency. The skill shells out to gh, which is the user's existing GitHub CLI, not a dependency this plugin installs.
• Evidence: a paper before/after check, described under Evaluation. Official wukong-code-evals was not run. This is not a claim of harness-wide proof.

What alternatives did you consider?

Putting the full type table only in AGENTS.md. That contradicts the request: plugin users read skills. The pointer in CLAUDE.md is the smaller overlap with the existing contributor habit.

A separate plugin. The behavior is part of using Wukong Code in any repo, and it adds no third-party dependency, so core is the distribution the request asked for.

Auto-close and Project board steps inside the skill. Those were not requested. The skill tells the agent not to do either.

Does this PR contain multiple unrelated changes?

No. One change: the recording-qa-to-issues skill. The CLAUDE.md pointer exists so the contributor guide does not contradict the skill. The README count exists so the catalog does not keep saying 26 skills after this one is added. skills/grilling/SKILL.md is untouched.

Existing PRs

• I have reviewed all open AND closed PRs for duplicates or prior art
• Related PRs: docs: add GitHub issue and Projects collaboration rules #50 (merged). No open PR. No closed PR adds this skill.

#50 added the lighter issue and Projects rules to CLAUDE.md. It did not add a skill, the 12-type table, or Conventional Commits issue titles. This PR does not replace #50. Closed issue #49 is grilling consensus tracking, which is a different workflow.

Environment tested

Harness (e.g. Claude Code, Cursor)	Harness version	Model	Model version/ID
Cursor	Cursor cloud agent, run bc-dd8704f9-06b0-5e1d-a246-762e57950385	Grok	grok-4.7

The skill text was not executed inside Claude Code, Codex, Kimi, OpenCode, or Pi.

New harness support (required if this PR adds a new harness)

This PR does not add a harness. No clean-session transcript applies.

Clean-session transcript for "Let's make a react todo list"

Not applicable. This PR does not add harness support.

Evaluation

• Initial prompt: 悟空非空也g asked to add a skill that records each agent-tool question and the reply as a new GitHub issue, with the type/label/title rules in the skill, not only in AGENTS.md.
• Official eval sessions after the change: 0. evals/ is not checked out here, and wukong-code-evals has no scenario for this skill. These notes are a paper behavior check, not that harness.
• Paper probes in this session: 1 baseline without the skill, then 2 with the skill loaded.
  • Baseline (no skill): title was the freeform sentence 「展开状态下点关闭后面板白屏，按钮太挤」, label bug only, and the rule would live only in AGENTS.md.
  • With the skill, same case: title fix(pane): 修复展开状态下关闭面板白屏, labels bug and ui, missing ui created first, a later message comments on the existing issue, and the rule stays in the skill.
  • Second with-skill case: the words "session token" stay in the question; a concrete secret value is replaced with [REDACTED].
• Not claimed: repeated adversarial sessions, or that every harness auto-triggers the skill.

Language-pack evidence

• Language and support status: none. This is not a language pack.
• Real failure generic methodology handled incorrectly: none
• Project markers and toolchain versions: none
• Official sources: none
• No-guidance repetitions and failures: none
• With-guidance repetitions and results: none
• Monorepo, unsupported-language, and docs-only controls: none
• Human reviewer familiar with the language: none
• Frameworks or third-party preferences introduced: none

none

Rigor

• If this is a skills change: I used wukong-code:writing-skills and completed adversarial pressure testing (paste results below)
• This change was tested adversarially, not just on the happy path
• I did not modify carefully-tuned content (Red Flags table, rationalizations, "human partner" language) without extensive evals showing the change is an improvement

skills/writing-skills/SKILL.md was read before writing. The paper check above is the evidence. It is not the adversarial multi-session campaign that checkbox describes, so those two boxes stay unchecked. No existing skill's Red Flags table or "human partner" wording was edited. The new skill has its own table.

Human review

• A human has reviewed the COMPLETE proposed diff before submission

悟空非空也g asked to open this PR. This session does not contain a review of the complete diff. The checkbox stays unchecked for that reason. Do not merge.