follow-up

Follow-up Status Check

The read-only front door onto this project's backlog. follow-up-triage's flow is the action step once you already know a follow-up needs promoting or closing; this skill runs first to find out whether one does, and -- because the follow-up list itself is often empty once things have been promoted -- to say what to work on next even when it is. See Backlog Triage and Ranking for the full, self-contained decision logic this skill applies.

Mindset

A follow-up sitting ACTIVE is a weak signal by design: someone flagged it and moved on, which is not the same as it still being the right thing to do next. TYPICALLY the useful answer to "what's open" spans three sources at once (the follow-up list, the register, the tech-debt list), not just the one the command name suggests. PREFER checking all three and saying which source the recommendation came from, over answering only the literal follow-up list and calling it done -- an empty follow-up list is a completely normal outcome (every follow-up promoted and closed) and never grounds for stopping the check early.

Rules

1. This skill never mutates the register, the tech-debt list, or a follow-up's status. Report and recommend only. Apply follow-up-triage's flow directly to promote or close a follow-up (it's an instructions doc, not a separate Skill), or hand off to the risk-register / tech-debt skills to file or resolve a row, or plan-create to scaffold work on the recommended item -- and only after the user says to proceed.
2. Zero ACTIVE follow-ups is not "nothing to do." It usually means the funnel is empty because everything in it was already promoted or closed -- the real backlog then lives in the register/tech-debt Open rows and the context index's DRAFT/ACTIVE PLAN/FINDING/KNOWN_ISSUE entries. Always fall through to those sources rather than reporting an empty follow-up list as the whole answer.
3. Ground the "next best item" pick in the value-rubric's read protocol wherever it applies (PLAN/FINDING/ KNOWN_ISSUE entries carry a value grade and a defined sort). For a pick sourced from the register or the tech-debt list instead, say explicitly that it's a qualitative judgment call -- those rows carry no value field, so there is no rubric-sorted answer to defer to.
4. If the context index looks stale, regenerate it (context-index skill) before trusting it. A quick sanity check: its header entry count should roughly match the actual file count under the context directory (see Prerequisites for the exact path). A mismatch means new or removed files haven't been indexed yet, and the candidate-pool step of the Workflow would be working from stale data.
5. An ACTIVE follow-up outranks a silently-pending backlog row. If the follow-up sweep finds any ACTIVE entry, lead the recommendation with it rather than the value-rubric or register/tech-debt pools -- a follow-up is something a person already flagged as worth acting on, which is a stronger signal than an unflagged row sitting in a list.