diff options
| -rw-r--r-- | .ai/sessions/2026-05-21-20-44-task-review-quick-tag-and-skill-install.org | 52 | ||||
| -rw-r--r-- | inbox/lint-followups.org | 5 |
2 files changed, 57 insertions, 0 deletions
diff --git a/.ai/sessions/2026-05-21-20-44-task-review-quick-tag-and-skill-install.org b/.ai/sessions/2026-05-21-20-44-task-review-quick-tag-and-skill-install.org new file mode 100644 index 0000000..ca8a166 --- /dev/null +++ b/.ai/sessions/2026-05-21-20-44-task-review-quick-tag-and-skill-install.org @@ -0,0 +1,52 @@ +#+TITLE: Session Context +#+DATE: 2026-05-21 + +* Summary + +** Active Goal + +Short maintenance session. Install the carried-over =review-code= skill so =/review-code= is routable, then add a =:quick:= effort-tagging rule to the task-review workflow at Craig's request. + +** Decisions + +- *=:quick:= is a per-task assessment during the task-review walk, orthogonal to the chosen action* — a task can be kept (or re-graded, or marked DOING) and tagged in the same pass; skipped on a Kill. Threshold is <=30 min, and the rule is *ask Craig when the heading/body don't make the effort clear* rather than guess, because a mislabeled =:quick:= wastes the spare-moment grab the tag exists for. +- *Edit the canonical =claude-templates/.ai/workflows/task-review.org=, not just the project copy.* =.ai/workflows/= is rsync-synced from the canonical with =--delete=, so a project-only edit reverts at next startup. Edited canonical + synced the mirror, verified identical. +- *Committed without the inline approval gate per Craig's explicit instruction*, but still ran =/review-code --staged= (verdict: Approve, no issues) and =/voice personal= (39 patterns). + +** Data Collected / Findings + +- =review-code= skill source existed at =review-code/SKILL.md= but had no symlink in =~/.claude/skills/=; =make install= is idempotent (skips already-linked, links only the missing) so it was safe to run. It also picked up two unlinked rules: =triggers.md= and =working-files.md=. +- =make doctor=: 40 ok / 0 warnings / 0 failures; no dangling symlinks. review-code now appears in the session's available-skills list. +- Inbox =lint-followups.org= held two non-actionable entries: a line-2070 =misplaced-heading= lint warning (verified false positive — =** Startup Pull Ordering= inside =...== verbatim markers in a DONE task body, same pattern as prior 2139/2143 cases) and the duplicate "12 tasks >30 days" health note. Not cleared this session — left for Craig's call. +- =/voice personal= fired #32 (first-person), #13 (em-dash split), #27 (passive), #34 (contraction); #39 scope check clean (claude-templates/rsync references are the change's own mechanics in a personal repo, not a leak). + +** Files Modified + +- =62d91e3 feat(workflows): tag <=30min tasks :quick: during task review= — pushed to =origin/main=. Three edits to the canonical workflow (+ synced mirror): new Phase C "Tagging =:quick:=" subsection, Phase D close-out mention, Common Mistake #7 (don't guess — ask). +- System-side (no repo change): =review-code=, =triggers.md=, =working-files.md= symlinked into =~/.claude/= via =make install=. + +** Next Steps + +- Inbox =lint-followups.org= still sits with the two non-actionable entries — clear when convenient (it regenerates from lint/wrap-up automation anyway). +- =docs/design/task-review.org= wasn't updated to record the =:quick:= action; add a one-line note there if the design doc should stay in step. +- Open =[#A]= carryover unchanged: =DOING= memory-sync (VERIFY on stow approach), =create-documentation= skill, =/update-skills= skill, 2026-05-04 audit review pass. +- Task-review nudge still standing: 12 top-level tasks unreviewed >7 days — a task-review cycle drops the count. + +* Session Log + +** Startup + review-code install + +Ran startup. Fresh session (no prior session-context). Installed the missing =review-code= skill via =make install= (also picked up two unlinked rules: =triggers.md=, =working-files.md=). =make doctor= 40 ok / 0 / 0. System-side only — no repo change. + +** Add =:quick:= tagging to task-review workflow + +Craig: any task estimated at <= 30 min should be tagged =:quick:= during a task review; ask him when the effort isn't clear rather than guessing. + +Edited the canonical =claude-templates/.ai/workflows/task-review.org= (the project =.ai/workflows/= copy is rsync-synced from it with =--delete=, so editing only the project copy would revert at next startup). Three edits: +- New Phase C subsection "Tagging =:quick:= — small tasks": estimate effort per task during the walk; tag <=30min, ask if unsure; orthogonal to the chosen action; skip on a Kill; tags share one cluster on the heading per todo-format.md. +- Phase D close-out summary now mentions newly-tagged =:quick:=. +- New Common Mistake #7: don't guess a =:quick:= estimate — ask. + +Synced canonical → project copy; verified identical. + +Committed + pushed per Craig's instruction (no approval gate, ran /review-code --staged + /voice personal). =/review-code= verdict: Approve, no issues. =/voice personal= fired #32/#13/#27/#34; #39 clean. =62d91e3 feat(workflows): tag <=30min tasks :quick: during task review= on origin/main (0/0 vs upstream). session-context.org left untracked, not part of the commit. diff --git a/inbox/lint-followups.org b/inbox/lint-followups.org index 5cce15e..42b0eb4 100644 --- a/inbox/lint-followups.org +++ b/inbox/lint-followups.org @@ -3,3 +3,8 @@ ** TODO line 2070 — misplaced-heading — Possibly misplaced heading line * 2026-05-20 Wed — Task-review health: 12 top-level [#A]/[#B]/[#C] tasks unreviewed for >30 days (daily review may have slipped) + +* 2026-05-21 lint-org follow-ups — todo.org +** TODO line 2070 — misplaced-heading — Possibly misplaced heading line + +* 2026-05-21 Thu — Task-review health: 12 top-level [#A]/[#B]/[#C] tasks unreviewed for >30 days (daily review may have slipped) |