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 agentmods add commands/dwarvesf/dwarves-kit/megagit clone --depth 1 https://github.com/dwarvesf/dwarves-kitWhat 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 | $0.00060 | $0.09751 |
| Opus 5 | $0.00030 | $0.04875 |
| Sonnet 5 | $0.00012 | $0.01950 |
| Haiku 4.5 | $0.00006 | $0.00975 |
Grade A, and why
mega 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 yesterday.
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 — 539 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a mega-goal planning lead. Your job is to turn ONE destination reached
through 3-8 DEPENDENT sub-goals into a committed roadmap on disk, then hand off to
whatever activator is present (/goal, ralph-loop, or the kit's own
lib/queue/orchestrate.sh). You do NOT implement anything yourself, and past the planning
step you never merge a PR whose ship-gate has not passed.
This is the kit-side half of the mega lane (ADR-0028 P2/P3, kit-hardening SG-08). It
mirrors the ops-toolkit plan-for-mega-goal skill's authoring beats -- decompose
(including the tiny-item batching rule), front-load every sub-goal's clarification
question ONCE up front, set the per-run merge config, and (as its own mode) Consolidate
overlapping same-project megas -- so a bare-kit team gets the same shape without the
skill installed.
It does NOT fork the skill: the skill stays the deeper authoring reference
(stacking-tool detection, scaffold-path resolution across arbitrary repos, the full
pointer-prompt convention, NOTES.md/FEEDBACK.md); this command is the kit-native
projection of the same three beats, scoped to the kit's own conventions
(lib/queue/orchestrate.sh's scaffold shape, lib/gate/gate-ledger.sh's lanes,
lib/gate/proof-ledger.sh's deployable classifier). Where the ops-toolkit skill is
installed, prefer it for anything this command does not cover; the two must never
diverge in checkpoint semantics -- a drift here is a bug in this file, not a feature.
Sibling to /kit:dispatch (SPEC-032): dispatch is the INDEPENDENT/parallel case
behind a disjointness gate; /kit:mega is the DEPENDENT/sequenced case -- one
bounded loop, one PR per sub-goal (SPEC-034). A real dependency GRAPH (fan-in,
fan-out, waves, topological order) is neither; that is the GSD v2 handoff tripwire,
not a reason to grow a scheduler here.
The intake ladder below (mirrors the skill's ladder verbatim, SPEC-142) is the FIRST check, before Prerequisites: it decides whether this command is even the right one to invoke for the task at hand.
Intake triage ladder (check first, before drafting anything)
Before drafting a goal, a scaffold, or a plan, classify the task against three rungs. This is a MUST-check routing step, not advice: a task that fits a lower rung never gets a higher rung's ceremony.
- DIRECT kit lane, in-session. One file, one behavior, one obvious proof.
lane-classifycalls it tiny or small: one worker drafts the change, verifies it, opens one PR. No scaffold, no conductor; the gate-ledger still records. Worked example: fixing a broken link in one README section, one file changed, verified by rendering the link. - Single
/goal. One objective, multiple steps, one stopping condition. Worked example: "users can sort the trade log by realized PnL", a multi-step feature with one verification command, handed to/goalviaplan-for-goal. - Mega-goal. Multiple objectives, multiple repos, or multiple gates converging on one destination. Worked example: the runner-fastpath mega-goal itself, eight sub-goals across three repos (dotfiles, dwarves-kit, ops-toolkit) each running its own
/goalwith its own PR, converging on one overnight runner.
Escape hatch: the user explicitly asking for a mega-goal overrides the ladder; never downgrade a request the user has already sized.
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.
- yesterday First seen · 539 lines · 60 tokens per session scan A d214b6a9c797
mega is a command published in the GitHub repository dwarvesf/dwarves-kit (11 stars, last pushed yesterday), licensed MIT. It adds 60 tokens to every session and 9,751 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 commands, from other repositories
OPSX: Archive
Archive a completed change in the experimental workflow.
OPSX: Explore
Enter explore mode - think through ideas, investigate problems, clarify requirements.
OPSX: Sync
Sync delta specs from a change to main specs.
OPSX: Propose
Propose a new change - create it and generate all artifacts in one step.
OPSX: Apply
Implement tasks from an OpenSpec change (Experimental).
OPSX: Update
Update a change - revise existing planning artifacts and keep them coherent (Experimental).