Problem
The maintainer named the actual current bottleneck in scoping #8103: "if my issues aren't extremely explicit, I usually end up finding more gaps/problems after merging their work." #8103 cuts the time cost of writing an explicit issue, but has no way to get BETTER at it over time — every draft is independent, with no memory of past misses. This issue closes that loop.
Area
Extends #8103's core module — not a separate tool.
Proposal
Deliverables
Resources
Boundaries
Cannot be built before #8103 exists. Does not attempt to auto-detect gaps from a merged PR — a human still decides what counts as a real miss.
maintainer-only — internal tooling quality loop, not a build task for contributors.
Problem
The maintainer named the actual current bottleneck in scoping #8103: "if my issues aren't extremely explicit, I usually end up finding more gaps/problems after merging their work." #8103 cuts the time cost of writing an explicit issue, but has no way to get BETTER at it over time — every draft is independent, with no memory of past misses. This issue closes that loop.
Area
Extends #8103's core module — not a separate tool.
Proposal
.claude/skills/contributor-pipeline-gardening/reference.md's own "what broke once already" sections). A more capable version could use the misses as few-shot examples in the drafting prompt itself — leave the exact mechanism to whoever builds this, informed by how tooling: maintainer CLI to expand a loose issue prompt into a gate-ready contributor issue #8103 actually turns out.Deliverables
Resources
.claude/skills/contributor-pipeline-gardening/reference.md(the existing style of "concrete lesson from a real incident" this should follow — see its "gate only enforces explicit text" and "reuse-existing-pattern" sections for the exact tone)Boundaries
Cannot be built before #8103 exists. Does not attempt to auto-detect gaps from a merged PR — a human still decides what counts as a real miss.
maintainer-only — internal tooling quality loop, not a build task for contributors.