docs: document contribution evidence gate and label taxonomy - #4478
docs: document contribution evidence gate and label taxonomy#4478mnriem wants to merge 4 commits into
Conversation
Add an "Evidence gate" subsection, a high-level "Review rubric" (D1-D7, scored 0-2 each), and a "Triage and author labels" reference to CONTRIBUTING.md so the acceptance bar is discoverable before contributing. Cross-links the existing AI-contributions section and three-open-PR cap rather than restating them. Assisted-by: GitHub Copilot (model: Claude Opus 4.8, autonomous) Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
There was a problem hiding this comment.
Copilot review overview
🟢 Approval recommended
The changes are documentation-only and appear consistent and well-structured, with only minor wording/typo nits noted inline.
Review tier: Lite
Findings: 1
New issues introduced by this change (1)
| Severity | Finding |
|---|---|
CONTRIBUTING.md — The phrase "hands off to" is ungrammatical here and reads like a typo; it’s likely meant to… |
What changed in this PR
This PR documents Spec Kit’s “evidence gate” and related review/triage guidance directly in CONTRIBUTING.md, making contribution expectations discoverable to contributors before they submit changes.
Changes:
- Adds an “Evidence gate” subsection describing what qualifies as acceptable evidence for changes.
- Adds a concise D1–D7 “Review rubric” table used to guide prioritization.
- Documents
triage-*verdict labels andauthor-*action labels, plus a link to an explicitly non-policy perspective piece.
| File | Description |
|---|---|
| CONTRIBUTING.md | Adds new PR-submission subsections for evidence, rubric scoring, and triage/author label taxonomy. |
Suppressed comments (1)
CONTRIBUTING.md:112
- "AI-source contributions" reads like a typo/awkward phrasing; "AI-sourced contributions" is the common grammatical form for contributions originating from AI output.
For further reading on the thinking behind this gate, see [one maintainer's perspective on AI-source contributions](https://blog.manorrock.com/blog/2026/09/08/spec_kit_ai_source.html). That piece is a personal viewpoint, not project policy — the policy is what's documented here.
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
Add author-needs-proof and author-needs-rebase to the author-actions table,
a "Community catalog submissions" subsection pointing catalog changes at the
submission-issue flow, and a note on when items are closed as
triage-out-of-scope. Fix the author-awaiting phrasing ("handed off to").
author-needs-rescope is intentionally omitted: it does not yet exist on
github/spec-kit, so it is flagged for the maintainer rather than documented
as a live label.
Assisted-by: GitHub Copilot (model: Claude Opus 4.8, autonomous)
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Assisted-by: GitHub Copilot (model: Claude Opus 4.8, autonomous) Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
There was a problem hiding this comment.
Copilot review overview
🟢 Approval recommended
The change is documentation-only and appears internally consistent, with only a minor wording nit identified.
Review tier: Lite
Findings: 1
New issues introduced by this change (1)
| Severity | Finding |
|---|---|
CONTRIBUTING.md — Link text uses "AI-source contributions", which reads like a typo/grammar error and is inconsistent… |
Issues resolved since last review (1)
| Severity | Finding |
|---|---|
CONTRIBUTING.md — The phrase "hands off to" is ungrammatical here and reads like a typo; it’s likely meant to… View resolved comment |
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
There was a problem hiding this comment.
Copilot review overview
🔵 Needs a closer look
The documented author-* label taxonomy currently doesn’t match the repo’s actual labels/descriptions (including a non-existent label), so the guidance would be misleading.
Review tier: Lite
Findings: None
Issues resolved since last review (1)
| Severity | Finding |
|---|---|
CONTRIBUTING.md — Link text uses "AI-source contributions", which reads like a typo/grammar error and is inconsistent… View resolved comment |
Suppressed comments (1)
Previously missed (1) — in code that hasn't changed since the last review.
CONTRIBUTING.md:113
- The “Author actions” label table doesn’t match the repository’s actual labels:
author-needs-rescopedoesn’t exist, and several descriptions differ from the current label descriptions (e.g.,author-needs-proof,author-needs-rebase,author-awaiting). To keep this section self-documenting and avoid drift, align the table to the exact label set/descriptions (or drop labels that aren’t present).

What & why
Spec Kit's acceptance bar — an "evidence gate" applied identically to human and AI-assisted contributions — currently lives in maintainers' heads. This PR moves it into
CONTRIBUTING.mdso the bar is discoverable before someone contributes. Docs/policy only; no CLI, tests, templates, or extensions are touched.Changes
Three new
###subsections under## Submitting a pull request(after the three-open-PR-cap paragraph, before### Branch naming):triage-can-waitand held behind proven work. Spells out what counts as evidence (reproduction/linked report, a regression test that fails onmainand passes with the change, scope discipline, disclosed AI use).0–2(14 max). Framed as guiding prioritization, not a hard pass/fail gate.triage-*verdict andauthor-*action labels, using the exact names/descriptions already on the repo.Existing content (the AI-contributions section and the three-open-PR cap) is cross-linked, not restated or contradicted. A companion perspective piece is linked as further reading and explicitly marked as a personal viewpoint, not policy.
Validation
markdownlint-cli2 CONTRIBUTING.md— no new violations introduced (the two pre-existing MD048 findings on the reporting-template blocks are unrelated to this change).gh label list --repo github/spec-kit.AI disclosure
This PR was authored by GitHub Copilot (model: Claude Opus 4.8) operating autonomously on behalf of @mnriem.