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-orchestrategit 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 | $0.00072 | $0.02343 |
| Opus 5 | $0.00036 | $0.01171 |
| Sonnet 5 | $0.00014 | $0.00469 |
| Haiku 4.5 | $0.00007 | $0.00234 |
Grade A, and why
v-orchestrate 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 — 65 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are running /v:orchestrate — the materialize step of Compound V. You turn a verified plan into the machine-readable contract the dispatcher runs: a manifest.yaml plus an initial state.json in a fresh run directory. This command does not dispatch — it produces the manifest and stops, so the user (or /v:dispatch) can inspect or edit it first.
The argument is {{args}}. It is normally a plan path. It may instead be an accepted fast-path pre-eval — a pre_eval_id (YYYY-MM-DDThhmmssZ-<slug>-<nonce>) or a path to docs/superpowers/pre-eval/<pre_eval_id>.json whose decision is FASTPATH_ELIGIBLE — in which case take Step 0 and stop.
The manifest schema and rules are defined in skills/compound-v/execution-manifest.md; the routing decisions come from skills/compound-v/routing-policy.md; the run-dir layout and state.json shape come from skills/compound-v/state-machine.md. Read those — they are the authority; this command is the procedure.
Steps
-
Fast-path branch (accepted pre-eval → committed single-job run). If
{{args}}resolves to an acceptedFASTPATH_ELIGIBLEpre-eval record (apre_eval_id, or adocs/superpowers/pre-eval/<pre_eval_id>.jsonpath withdecision: FASTPATH_ELIGIBLE), do not run the plan-based flow below — the fast path has no full plan and no three audits. Instead delegate to the deterministic materializer, which runs the authoritative Phase-M lifecycle (mint a deterministic run-id frompre_eval_id→ copy the pinned taxonomy snapshot into the run → write spec/plan stubs, block-YAML audit skip-records, the single-jobfast_pathmanifest with the review declaration only, and the captured implementer prompt → commit all artifacts exceptstate.json→ append + commit thebindevent → commitstate.jsonatFASTPATH_DISPATCHEDLAST). It also runs the validator in--mode pre-dispatchas an in-code gate before binding, so a manifest the validator would reject never reaches dispatch.python3 scripts/compound-v-fastpath-materialize.py materialize \ --repo . --pre-eval-id <pre_eval_id> [--prompt-file <captured-implementer-prompt>]The materializer is idempotent + crash-consistent: a committed
state.jsonatFASTPATH_DISPATCHED⇒ thebindis already durable ⇒ the run is complete (re-running is a no-op); a run interrupted beforestate.jsonis rebuilt deterministically (samepre_eval_id→ same run-id; an existing child is discovered before another is minted). It fails closed — a tampered record (itslocalization.resolved_paths[0]disagreeing with the committed localization artifact, a non-FASTPATH_ELIGIBLEdecision, or a taxonomy snapshot whose bytes don't content-address to the record'staxonomy_digest) is rejected before any write or commit. Report the run-id and the next step (/v:dispatch <run-id>), then stop — Steps 1-9 below do not apply. -
Resolve the plan. If
{{args}}is empty, list the plans indocs/superpowers/plans/and ask the user which to materialize. If the named plan path does not exist, stop and say so. -
Read the plan and its pre-flight audits. Extract:
- the feature title and feature-level Acceptance Criteria (from the spec the plan targets),
- the verified Partition Map (the disjoint write-paths per task),
- the three audit paths (
docs/superpowers/{archaeology,expert,library-audit}/<topic>.md). If the plan has no Partition Map, stop — this plan is not Compound-V-ready; route it through Phase 2 (skills/compound-v/phase-2-disjoint-partitioning.md) first.
-
Choose the run-id. Convention:
YYYY-MM-DD-<slug>(the slug from the plan/feature). The run dir isdocs/superpowers/execution/<run-id>/. Create it (andjobs/+results/subdirs). -
Apply the routing policy (
routing-policy.md) to each partition entry. For every job setbackend · model · isolation · runfrom itstype, honoring the active stance in.claude/compound-v.json(Balanced default; Claude-only when Codex is absent). Consultdocs/superpowers/memory/routing-lessons.md— a recorded lesson overrides the table default for that pattern.
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 · 65 lines · 72 tokens per session scan A edcda01c1ef4
v-orchestrate is a command published in the GitHub repository procoders/superpowers-v (35 stars, last pushed 3d ago), licensed MIT. It adds 72 tokens to every session and 2,343 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
status
Show current project status — active feature, feature counts, git tree, TOC freshness, session handoff. Read-only overview. Distinct from /session-end (which writes).
strict_full_cycle
Interactive workflow for the entire development cycle with every review gate forced on, ignoring the reviews a repository disabled in its configuration.
gbu-status
Herd the running agents — who's idle, who owes a report, who to verify.
template
Manage issue templates for streamlined issue creation.
speckit.auto
Automatically execute the four core phases of the Spec-Driven Development (SDD) pipeline: specify → plan → tasks → implement, in strict sequential order.
OPSX: Bulk Archive
Archive multiple completed changes at once.