Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add phj128/autoworker --skill deep-plangit clone --depth 1 https://github.com/phj128/autoworkerWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/phj128/autoworker/deep-plan)<a href="https://agentmods.dev/skills/phj128/autoworker/deep-plan"><img src="https://agentmods.dev/badge/skills/phj128/autoworker/deep-plan.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00050 | $0.01848 |
| Opus 5 | $0.00025 | $0.00924 |
| Sonnet 5 | $0.00010 | $0.00370 |
| Haiku 4.5 | $0.00005 | $0.00185 |
Grade A, and why
deep-plan scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 185 lines — stays where its author put it; the contents beside it link to each section on GitHub.
autoworker:deep-plan — Structured Deep Discussion (Inside Plan Mode)
Invoke immediately after entering Plan Mode. Ensures discussion depth through 5 structured phases, producing a structured plan file for autoworker:subtask-init to extract from.
Core idea: Plan depth determines the quality ceiling of the execution chain. autoworker:subtask-init extracts from the plan file, autoworker:subtask-plan derives the verification plan from it. If the discussion is shallow, the entire chain — no matter how well-structured — is verifying an insufficiently thought-through solution.
Execution Flow
Phase 1: Motivation Exploration
Goal: Understand what problem the user truly needs to solve, rather than rushing to think about how.
Pre-action: Check if the project has a task_plan.md. If so, read the overall goals and phase list; subsequent questioning should address the relationship to the overall plan.
Questioning approach (use AskUserQuestion, 2-3 questions per round):
- Why do this? Do not accept tautological answers like "because we need X". Probe to root causes at the business/experience/efficiency level.
- Relationship to overall goal? (when task_plan exists) Which phase does this correspond to? Is it planned or ad-hoc? If ad-hoc, why is it higher priority than planned phases?
- What if we don't do it? What's the worst consequence? Can we tolerate it? The lighter the consequence, the simpler the solution should be.
- Is this really needed? Is there a simpler way (don't do it, do it manually, approach from a different angle)?
Anti-patterns:
- User says "add caching" → immediately designing cache solution → should first ask "why is it slow?"
- User says "refactor to X" → immediately designing X architecture → should first ask "what problem does refactoring solve?"
- User gives detailed solution → skip motivation and discuss implementation → the solution might be the user's guess, not optimal
Depth gate: Motivation expressible in 1-3 clear sentences (WHY + consequence of not doing it). Can't write it = don't understand it = keep asking.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 7d ago First seen · 185 lines · 50 tokens per session scan A 9ade6445045f
deep-plan is a skill published in the GitHub repository phj128/autoworker (18 stars, last pushed 5mo ago), licensed MIT. It adds 50 tokens to every session and 1,848 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
batch
Dispatch a night's batch of Todo tickets — one worktree and one pull request each — and leave a morning summary. Run explicitly; never inferred.
cleanup
Clean up after a merged pull request — return to the base branch, remove the worktree, delete the branch, sweep other stale branches and worktrees, and report tracker state. Use after a merge or whenever asked what is left over.
prep
Spec a ticket before it is dispatched. Brainstorms it, decides everything it can recommend an answer for, asks a present human about the little that survives, and leaves the spec, the decisions and anything still open as one ticket comment. No worktree, no branch, no pull request, and the ticket stays where it was.
work
Dispatch a tracker ticket into an isolated worktree and drive it autonomously to an open pull request, then stop for review. Pass interactive to brainstorm the ticket with a human before planning it. Use when asked to work a ticket end-to-end.
merge
Merge the reviewed pull request for the current branch, then hand straight off to cleanup. Use when a human green-lights an open PR.
daily-working
End-to-end pipeline: pull a task from Redmine by ID, sanity-check and impact-assess it against the codebase before touching anything, implement it with the Claude CLI, verify the result in a real browser via the Claude Chrome extension (claude-in-chrome), and keep the Redmine ticket in sync throughout (in-progress…