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/procoders/superpowers-v/v-dispatchgit clone --depth 1 https://github.com/procoders/superpowers-vWhat 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 | $0.00083 | $0.01351 |
| Opus 5 | $0.00042 | $0.00675 |
| Sonnet 5 | $0.00017 | $0.00270 |
| Haiku 4.5 | $0.00008 | $0.00135 |
Grade A, and why
v-dispatch 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 2d 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 — 56 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are about to execute Phase 3 of Compound V — batched parallel multi-backend dispatch — on {{args}}.
This replaces the default Superpowers subagent-driven-development sequential-implementer pattern with parallel batches (4–6 implementers per message) on Opus by default, Sonnet only where the routing policy's justification holds, and Codex for large isolated jobs — each job dispatched through backend-launcher and gated by the file-scope check on return.
{{args}} is accepted in three backward-compatible forms — the dispatcher detects which:
{{args}} is… |
Action |
|---|---|
a plan path (docs/superpowers/plans/…md) |
materialize the manifest first (Phase 2 → manifest.yaml + state.json in a new run dir), then dispatch. This is the 0.1.x contract the plan-saved-nudge hook and current users rely on — it still works. |
a manifest path (…/execution/<run-id>/manifest.yaml) |
dispatch it directly (already materialized). |
a run-id (a dir name under docs/superpowers/execution/) |
resolve to that run's manifest.yaml and dispatch directly. |
Steps
-
Resolve
{{args}}.- If
{{args}}is empty, list plans indocs/superpowers/plans/and runs indocs/superpowers/execution/, and ask which to dispatch. - If
{{args}}is a run-id or a manifest path, load that run'smanifest.yaml. Skip to step 3 (already materialized). - If
{{args}}is a plan path, verify it exists, then materialize per/v:orchestrate: applyrouting-policy.md, writemanifest.yaml+ initialstate.jsonintodocs/superpowers/execution/<run-id>/(schema:execution-manifest.md; run-dir + state shape:state-machine.md), then continue.
- If
-
Validate the materialized manifest (only for the plan-path branch). Pick the validator mode by manifest kind (CR5-1): if
manifest.yamlcarries afast_pathblock, validate it in pre-dispatch mode; a legacy (plan-based) manifest carries no such block and is validated mode-lessly, as before — a mode-lessfast_pathmanifest is fail-closed rejected:# legacy manifest (no fast_path block): python3 scripts/compound-v-validate-manifest.py docs/superpowers/execution/<run-id>/manifest.yaml # fast_path manifest (v2.9 pre-eval-backed): python3 scripts/compound-v-validate-manifest.py docs/superpowers/execution/<run-id>/manifest.yaml \ --mode pre-dispatch --repo-root <repo>Non-zero exit ⇒ fix the manifest and re-run; do not dispatch a manifest the validator rejects. (A plan-path materialization always yields a legacy manifest; the
fast_pathbranch matters when{{args}}is a manifest/run-id that resolves to a pre-eval-backed run — and the partition reviewer in step 3 re-runs the validator with the same mode rule.) -
Run the partition reviewer first (Iron Rule #4: no execution without a verified Partition Map):
- Dispatch
compound-v:partition-reviewerwith the plan and the manifest (it runscompound-v-validate-manifest.pyas its deterministic backing gate, then verifies disjointness + invariants). - If verdict is
FAIL→ STOP. Surface the failure to the user. Do not dispatch implementers. - If verdict is
PASS→ continue.
- Dispatch
-
Dispatch the parallel dispatcher:
- Dispatch
compound-v:parallel-dispatcherwith:- the manifest path (and run dir),
- the partition-review verdict (PASS),
- the audit paths:
docs/superpowers/{archaeology,expert,library-audit}/<topic>.md.
- The dispatcher handles Task 0 serially, then the parallel batches (honoring
depends_on/run/max_parallel), routing each job to itsbackendviabackend-launcher. After every job it runs the scope gate (git diff --name-only∪ls-files --othersvswrite_allowed): a BLOCKED job HALTS the run and is never merged. It updatesstate.jsonafter every phase, then runs the collector and the three-pass Review Gate (AC-gated).
- Dispatch
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.
- 2d ago First seen · 56 lines · 83 tokens per session scan A 59d552f2dcbe
v-dispatch is a command published in the GitHub repository procoders/superpowers-v (35 stars, last pushed 2d ago), licensed MIT. It adds 83 tokens to every session and 1,351 once invoked, about $0.0004 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
deepen
Spare-budget design pass. Make one shallow module deep — smaller interface, behavior held, tests green before & after.
research
Gather external knowledge into §R so build grounds in facts, not hallucinations. Every finding cites a source.
review
Adversarial senior review of the spec before build. Refute, don't rubber-stamp. Ends in a go/no-go gate.
build
Plan-then-execute against SPEC.md. Native Claude Code loop, no sub-agents.
check
Drift detector. Diff SPEC.md against code. Read-only, zero writes.
full_cycle
Interactive workflow to execute the entire development cycle from brainstorming to the integration decision.