whats-next

Use when asked "what should I work on next?", "what's the backlog?", or when planning a session, prioritizing a project's open work, or checking that nothing pending has been missed. Assumes the docs-

SKILL.md

GitHub

What's next (backlog triage)

Answer "what's next?" by sweeping every backlog source, separating live-actionable work from accepted constraints, and returning a categorized, prioritized backlog plus a coverage manifest. Prioritization is innate to a capable agent; the rules below force the two things unguided runs omit — the manifest ("did I see everything?") and the accepted-constraints list ("considered; by design; skip") — plus sweep completeness. This is the planning layer on top of the docs-architecture family (it consumes audit flags and maintain's residue as backlog inputs; doc hygiene itself belongs to the siblings). After an audit when doc currency is in doubt: stale docs → stale backlog, so currency-check first.

Sweep — every row; manifest whatever you can't reach

Source Yields
working tree + git status / log uncommitted WIP to finish; recent direction
ROADMAP "Future directions" / "Next" / known-limitations planned features + accepted limits
CLAUDE.md + docs gotchas hazards — most are accepted constraints, not work (W1–W2)
journal docs/ + NOTEBOOK.md "pending / open items" un-promoted decisions, open questions
code TODO / FIXME (note whose — first-party vs vendored/upstream) actionable only if first-party
planning/change-tool state (e.g. openspec/ changes, issue tracker) in-flight work
audit findings (docs-architecture-audit output) doc-drift fixes
cross-repo blockers / consumers dependencies + sequencing
tests / build status red = urgent

Rules

  • W1. Live-actionable requires a live marker: a first-party TODO/FIXME, an audit flag-code-bug, a ROADMAP "Next" item, or an explicit pending/open marker. Everything else — by-design, "do not re-litigate", deferred-by-decision, vendored/upstream — → the accepted-constraints list (shown, not actioned).
  • W2. Default to accepted when unsure. (A false "todo" on every gotcha trains the user to ignore the output.)
  • W3. Bucket each actionable: urgent (red tests / broken) · deferred-fix · future-feature · doc-debt.
  • W4. Rank by value × (1 / effort) × exposure-if-deferred — the cost of not doing it (what it blocks or de-risks); implementation riskiness alone never raises rank. Tag effort (XS/S/M/L) + risk.
  • W5. Deliver a top-N with a recommended sequence (what unblocks / de-risks what) — never a flat list. Priority is a proposal; the user owns the call.
  • W6. REQUIRED: reuse the fan-out from docs-architecture-audit — independent passes, merge, loop-until-dry. (Passes agree on the core; the margins are a coin-flip.)

Two REQUIRED outputs

  • W7. Coverage manifest — a table over every sweep row: swept (with what it yielded) or not reached (and why — no run logs, vendored, sibling repo). (Without it the backlog reads complete when it isn't — the #1 unguided omission.)
  • W8. Accepted constraints — "considered, not work" — each with the one-line reason to skip. (Shows the work; prevents re-litigation.)

Full rationale + RED/GREEN provenance: https://github.com/Apoplectic1/YetAnotherSkillsRepo/blob/main/docs/docs-architecture-design.md