Feat/router start here#6
Merged
Merged
Conversation
Make the skill a self-sufficient entry point. Until now the router only mapped intent to sub-skills; an agent that discovered/installed the skill (before ever seeing the website) had no in-skill cue to onboard the repo first. Add a 'Start here (first run)' section that mirrors deepworkplan.com/init.md, locally and network-free: if the repo is not yet AI-first (no root AGENTS.md / no .agents/), read the local spec + adaptation rule (REASON, do not copy-paste), run the non-destructive onboard flow (reconcile, ask before replacing), then create and execute Deep Work Plans. If already AI-first, skip onboarding and route by intent. Reinforces the model: the skill is the reusable engine; the repository is what gets adapted. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add a read-only 'verify' sub-skill (/dwp-verify) that checks, mechanically, whether a repository is DeepWorkPlan-conformant (AI-first) and whether its plans are well-formed — mirroring the spec's Conformance criteria: - repo: AGENTS.md with real runnable commands, CLAUDE.md resolution, docs/ categories, .agents/ + catalog-matches-disk, thin dwp-* delegators, gitignored .dwp/ (plans/ drafts/) + tmp/, resolvable skill; - plan: every task has acceptance criteria + a validation gate, persisted progress, the two mandatory final tasks, re-anchoring. Reports pass/fail (CONFORMANT / NOT CONFORMANT — N issues) and offers to capture gaps as a Deep Work Plan; never fixes silently. Wire it into the router (routing row + 'Start here' verify step + description), and have onboard generate the /dwp-verify delegator (six dwp-* commands now). validate-frontmatter: 12/12 OK. bats: 16 ok. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Make hours-long unattended execution an explicit, safe path: in trust/auto mode, loop task -> validate -> commit -> update PROGRESS.md -> next without per-task confirmation; stop only on a failed gate, ambiguity, a blocking dependency, an out-of-scope/destructive action, or completion. Emphasize that state lives on disk (README checkboxes, Completion & Logs, PROGRESS.md) so a plan survives a context-window reset, and that resume re-reads disk (never conversation memory) and re-anchors to the goal each task to prevent drift. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
PR: feat — router Start-here + verify sub-skill + execute autonomous mode
Repo:
DailybotHQ/deepworkplan-skillBase:
main← Head:feat/router-start-hereCreate: https://github.com/DailybotHQ/deepworkplan-skill/compare/main...feat/router-start-here?expand=1
Title:
Summary
Finalize the skill so any agent can take a repository from zero to fully
agent-pilotable — discover the skill, make the repo the harness, then create,
verify, and execute Deep Work Plans autonomously for hours. Three additions:
Router "Start here (first run)" — make the skill a self-sufficient entry
point. Until now the router only mapped intent → sub-skill; an agent that
discovered/installed the skill (before ever seeing the website) had no in-skill
cue to onboard. The router now mirrors
deepworkplan.com/init.mdlocally(no network): if the repo isn't AI-first (no root
AGENTS.md/ no.agents/),read the local spec + adaptation rule, run the non-destructive
onboardflow,verify, then plan/execute.
verifysub-skill (/dwp-verify) — a read-only, objective conformancecheck. Reports pass/fail for the repository (AGENTS.md with real runnable
commands, CLAUDE.md resolution, docs/ categories,
.agents/catalog matchingdisk, thin
dwp-*delegators, gitignored.dwp/+tmp/, resolvable skill)and for plans (every task has acceptance criteria + a validation gate,
persisted progress, mandatory final tasks, re-anchoring). Wired into the router
(routing row + Start-here step), and
onboardnow generates the/dwp-verifydelegator (six
dwp-*commands). Mirrors the website's/specConformance doc.executeautonomous mode + context-window checkpointing — make hours-longunattended runs explicit and safe: in
trust/auto, looptask → validate → commit → update
PROGRESS.md→ next, stopping only on afailed gate, ambiguity, a blocking dependency, an out-of-scope/destructive
action, or completion. State lives on disk so a plan survives a context-window
reset;
resumere-reads disk (never conversation memory) and re-anchors to thegoal each task to prevent drift.
Validation
python3 scripts/validate-frontmatter.py→ OK, all 12 SKILL.mdbats tests/→ 16 ok (1 skip: codex on PATH)Release / downstream
This is additive: a new sub-skill + router/onboard/execute docs → minor bump
(v2.2.0) via auto-release. After release, the deepworkplan-website pin
(
skills-lock.json) is updated to 2.2.0 so the site's "eight sub-skills +/dwp-verify" matches the published skill.
🤖 Generated with Claude Code