메뉴
Resume an interrupted Deep Work Plan from its recorded progress state. Use when the developer wants to continue a plan in .dwp/plans/ that was paused or interrupted mid-execution.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | deepworkplan-resume |
| description | Resume an interrupted Deep Work Plan from its recorded progress state. Use when the developer wants to continue a plan in .dwp/plans/ that was paused or interrupted mid-execution. |
| version | 2.15.0 |
| documentation_url | https://deepworkplan.com |
| user-invocable | true |
| allowed-tools | Bash, Read, Grep, Glob, Edit, Write |
Safely continue an interrupted plan from where it stopped — no duplicated work,
strict order, continuing from the first [ ] task.
../shared/context.sh — resolve repo root, branch, and
dwp_dir.../shared/dwp-paths.md — plans at
.dwp/plans/PLAN_{name}/.../shared/adaptation.md — the two repository
archetypes (relevant for orchestrator/child-DWP awareness).../execute/SKILL.md — the execution rules this flow
resumes into (team-agents, orchestrator task types, Dailybot golden rule)./dwp-resume {plan_name} — resume directly (skip the menu)./dwp-resume latest — resume the most recently modified plan.Normalize the PLAN_ prefix; validate .dwp/plans/PLAN_{name}/ and its
README.md. If not found, show available plans and ask the user to choose.
If a parameter was given, resolve the plan (or latest), validate folder +
README under .dwp/plans/, and skip to Step 2. Otherwise go to Step 1.
List PLAN_* folders in .dwp/plans/; mark the most recently modified as
latest. Present a numbered menu (number / name / latest) and validate the
choice.
This step implements the DWP Resume Protocol
(../spec/DWP_SPECIFICATION.md §5.3) — the named six-step ritual any resuming
session performs:
[x] vs [ ]).git status, git log --oneline -10, and git diff (if uncommitted
changes exist). In a workspace without git, read state.json's checkpoint
instead (../spec/PLAN_STATE.md §4.4).state.json against
the README checkboxes. On any desync the markdown wins — regenerate
state.json from the README (and git log), note the reconciliation in
PROGRESS.md, then continue (../spec/PLAN_STATE.md §5). If
state.json.blocked is set, surface it: that is why the plan stopped.[ ] task.git status shows uncommitted changes, review them
against the current [ ] task; read that task's Completion & Log for notes.AGENTS.md Quick Commands) to confirm the world still works before adding to
it. A failing smoke test is investigated first.Report: completed [x] tasks; pending [ ] tasks; the task to resume from; git
state (uncommitted changes, recent commits, last commit); the current task file
and any completion-log notes; partial-work findings; and the plan to finish the
task. Location: .dwp/plans/PLAN_{name}/.
If uncommitted changes exist, classify them: valid partial progress (incorporate and complete), incomplete (finish remaining work), or unclear/unrelated (ask the user before proceeding). If none, proceed directly.
[x] tasks — trust the README, cross-check git commits.[ ] tasks — strict order; complete each fully.[ ] task — open its file, read fully, address any
logged blocker first.[ ], report and wait;
on pass, update the log, mark [x], and commit.[ ] task until done or paused.From here, the standard Execute rules apply (
../execute/SKILL.md): the per-task significance Dailybot report, the plan-completion milestone golden rule, and — for orchestrator plans — orchestrator task types (create_child_dwp/integration_checkpoint/execute_child_dwp), manifest checks, and team-agents parallel groups (real team agents, not subagents) with sequential fallback. Resume from the first[ ]task using those same rules, including hub-vs-individual-repo navigation per../shared/adaptation.md.
If a task log shows a blocker, read it, fix the issue, re-run validations; on
pass: mark [x], commit, continue; on fail: log again, stop, request help.
After resuming and completing a task: ✓ Resumed and completed: Task N with
files changed, validation result, commit, and the next task. On blocker: ✗ Task N blocked with the blocker, action taken, status, and next step.
[x] done / [ ] pending); verify with git; read
completion logs; never duplicate work; never skip; assess partial work
carefully; stop on blockers.Create blog posts — from a topic (writes content) or with provided content (scaffolding). Use proactively when creating new blog posts or articles.
Pre-publication audit for blog posts — comprehensive final review of SEO, AEO, accessibility, images, content quality, i18n parity, and project conventions before publishing. Use proactively before publishing any blog post.
Pre-publication audit for blog series — validates series definition, post ordering, cross-post consistency, navigation, and runs individual post audits for all posts in the series. Use proactively before publishing any blog series.
Audit the blog tag taxonomy — frequency analysis, orphan detection, hierarchy validation, and proposals for new subtopic tags. Read-only — proposes, never modifies tags or posts. Use proactively before each release cycle or after a content drop of 5+ posts.
Optional DeepWorkPlan addon that connects an AI-first repo to the developer's Dailybot team — installing (with consent) the Dailybot agent skill (DailybotHQ/agent-skill) and/or the Dailybot CLI (DailybotHQ/cli), wiring the plan lifecycle into best-effort agent updates - kickoff when a plan starts, significant task completions, a blocked report when an unattended run halts, and a milestone on plan completion - with payloads derived from the plan's state layer, and optionally committing the Dailybot skill's deterministic hook enforcement (dailybot hook lifecycle hooks, CLI >= 1.12.0) so the agent harness itself reminds agents about unreported work. Opt-in, never required, never blocks the work, reconciles existing setups instead of clobbering them, and defers all auth to the Dailybot skill's own consent flow. Use when the developer or team already uses Dailybot and wants DWP progress visible to humans.
Optional DeepWorkPlan addon that safely upgrades a repo's dependencies — reasoning about the repo's ACTUAL package manager (npm/pnpm/yarn + ncu, pip/poetry/uv, cargo, go mod, bundler, composer, and more) rather than assuming npm — with a batched, validated, revertible workflow that detects the manager and manifests/lockfiles, classifies upgrades (patch/minor/major), upgrades in safe batches, runs the repo's real validation gate after each batch, reverts a failing batch, and summarizes. Opt-in, never required, reconciles with the repo's existing tooling. Use when the developer wants to bring dependencies up to date without breaking the build.