Skip to content

duthaho/skillhub / .claude/skills/revisit

revisit

Verified

Revisit dashboard — one scan of everything you parked to re-check later, across every skill's memory. Reads out/*/tracker.md and the other memory files skills leave behind (ideate PARK "revive when", verdict/scout revisit-when triggers, refactor WATCH candidates, learn spaced-repetition due dates, jobfit roles to re-check) and surfaces what's due now: date-based items past their date, trigger-based items whose condition may have fired. Renders one prioritized "due to re-check" list in chat — grouped by skill, most overdue first — quoting the exact tracker line and its revive trigger so you can act or defer. Strictly read-only: it never edits a tracker or fires the follow-up skill; it points, you decide. Use for "what do I need to revisit", "what's due to re-check", "anything parked worth reviving", "show my revisit queue" — e.g. "/revisit", "/revisit ideate".

revisit — what you parked, now due

/revisit — scan every skill's memory for items due to re-check /revisit <skill> — only that skill's tracker (e.g. /revisit ideate)

Skills that judge under uncertainty don't just decide — they record a trigger for revisiting the decision: ideate PARKs an idea "revive when X", verdict stamps "revisit-when", scout/refactor leave WATCH items, learn schedules the next review by date. Scattered across out/*/, those triggers rot unseen. This skill is the one place that gathers them and asks a single question: what is due to re-check right now?

It is strictly read-only — the same contract as daybrief. It never edits a tracker, never marks anything revived, never runs the follow-up skill. It surfaces; you decide.

Step 0 — Locate the memory

Today's date is date +%F (compute it; don't hardcode). Then find out/ in the current project.

  • No out/, or no memory files under it → say so plainly and stop. Nothing to revisit is a valid, complete answer (keyless, graceful — never invent items).
  • /revisit <skill> → restrict the scan to out/<skill>/ only.

Completion criterion: the set of memory files to scan is fixed, or a clean "nothing to revisit yet" is reported.

Step 1 — Collect the revisit signals

Read each memory file and pull every line carrying a revisit signal. Known shapes (don't limit to these — any "revive/revisit/recheck/due" marker counts):

SourceSignal to pull
out/ideate/tracker.mdPARK rows with a "Revive when …" reason
out/verdict/*.mdrevisit-when triggers on a past verdict
out/scout/tracker.mdWATCH items and their revive condition
out/refactor/tracker.mdWATCH candidates deferred to a trigger
out/learn/*.mdnext-review due dates (spaced repetition)
out/jobfit/tracker.mdroles marked to re-check later

Skip anything already resolved (SHIPPED, DONE, DROP, applied) — a DROP is not a revisit. Capture for each item: source skill, the verbatim line, and its trigger (a date, or a condition in prose).

Completion criterion: a flat list of open revisit items, each tagged with source skill and its trigger.

Step 2 — Classify each item

Sort every item into exactly one bucket:

  • DUE — a date-based trigger whose date is on or before today, or a condition-based trigger that plausibly has fired given what you can see in the repo/trackers (state the evidence). When unsure whether a condition fired, mark DUE and say why it's worth a look — a false "check this" is cheaper than a missed revive.
  • NOT YET — date still in the future, or condition clearly not met. Keep it, but below the fold, with the unmet trigger shown.

Don't fire or resolve anything — classification only.

Completion criterion: every item is DUE or NOT YET, each with a one-line reason.

Step 3 — Render the queue

One prioritized list in chat:

  • DUE first, most overdue at the top (oldest date, or longest-standing condition). Group by source skill. For each: the quoted line, its trigger, why it's due, and the next skill to run if the user chooses to act (e.g. "→ /ideate to re-diverge", "→ /verdict revisit <brief>").
  • NOT YET collapsed underneath as a short "watching" list — item + the trigger not yet met — so the user sees the horizon without noise.
  • Nothing DUE → say so in one line; the watching list is still worth showing.

Never auto-run a follow-up. Naming the next skill is a suggestion the user acts on — the read-only contract holds to the last line.

Completion criterion: a single grouped, prioritized queue in chat; every DUE item names the skill that would act on it; nothing was written or triggered.

reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 469kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 321kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 321kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 321kUpdated Claude CodeCodexVerifiedInstall