hyperflow
Advanced multi-agent orchestration with persistent cross-session memory, per-step multi-level review, persona stitching, and adaptive flow profiles
Installation
Open Claude Code and run this command:
/plugin install hyperflow@claude-code-plugins-plus
Use --global to install for all projects, or --project for current project only.
What It Does
Hyperflow
Multi-agent orchestration for Codex App/CLI, Claude Code, OpenCode & Antigravity.
Thinking models plan and review every step. Worker models execute in parallel. Learnings persist in local, per-project memory.
amplify → spec → scope → dispatch → audit → deploy
Start anywhere. Auto-advance forward. Memory persists across sessions.
Skills (27)
Use when a prompt is rough, vague, or under-specified and you want it rewritten to high quality before running it.
Amplify
Turn a rough prompt into a high-signal one. Detect the domain, inject the standards that domain demands, score it against a rubric — then offer to run it.
Dispatcher and reviewer — Opus 4.8 (thinking-tier). Searcher/Writer — Sonnet 4.6 (worker-tier).
Amplify does one thing well: it rewrites the prompt you give it into the single strongest version, then hands it off. It does not write code itself — it produces a prompt other skills (or you) execute.
Per-Step Agent Map (DOCTRINE rule 12)
Every substantive step dispatches at least one Agent. Atomic steps (DOCTRINE 12.2.8) are a single Worker → Reviewer pair with no independent angles to fan out.
| Step | Status | Worker tier | Thinking tier | Notes |
|---|---|---|---|---|
| 1 — Read intent | Atomic (12.2.8) | Searcher (Sonnet) — domain signals + project rules | Analyst (Opus) — triage + gap analysis + persona selection | Single Worker → Reviewer; reads CLAUDE.md, AGENTS.md, .hyperflow/memory/* |
| 2 — Amplify | Atomic (12.2.8) | Writer (Sonnet) — draft the enhanced prompt | Reviewer (Opus) — rubric score, one targeted revision | Writer drafts → Opus scores against the 8-dim rubric → revise once if any dim < 4 |
| 3 — Present + handoff | §12.1 inline | — | — | Print the amplified prompt + rationale; fire the handoff gate (AskUserQuestion) |
Inputs
$ARGUMENTS— the raw prompt to amplify. If empty, amplify the user's previous message in the conversation.- No flags. Amplify always produces the single best version (per the design decision); the rubric, not a flag, governs depth.
Flow
Step 1 — Read intent (atomic · 12.2.8)
Atomic — a single Searcher → Analyst pair. No parallel angles: understanding one prompt is one scope.
- Dispatch
Searcher — detecting prompt domain + gathering project rules(Sonnet). It returns:
- Domain signals — does the prompt concern frontend/ui/creative, api/db backend, mobile, security, performance, refactor/bugfix/test, devops, docs? (A prompt can span several.)
- Project rules — read, when present:
CLAUDE.md,AGENTS.md,.hyperflow/memory/conventions.md,.hyperflow/memory/project-decisions.md,.hyperflow/memory/anti-patterns.md. Extract any rule that should constrain the prompt. - The raw prompt's gaps — what a senior engineer would have to guess to act on it.
- Dispatch
Analyst — triaging intent + selecting personas(Opus). It returns{
Use when the user wants a code review on recent changes — quality, spec, security, or performance feedback.
Audit
Multi-level code review. Dispatcher — Opus 4.8 (thinking-tier). Workers — Sonnet 4.6.
This skill exercises Layer 3 (Orchestrator) and Layer 9 (Security). After the review prints, a fix gate asks the user whether to apply the findings — on Yes, audit auto-invokes /hyperflow:scope with the findings as the spec, which then chains to /hyperflow:dispatch.
Iron Rules
Failure recovery (DOCTRINE rule 14). Worker errors, malformed output, NEEDS_REVISION verdicts, and gate failures in every Step follow the canonical policy in skills/hyperflow/failure-recovery.md. Audit-specific exception: a failed Reviewer at L1/L2 escalates to an L3+ Reviewer at the same severity level rather than aborting — audit exists to catch issues, so a Reviewer failure is best resolved by a more thorough Reviewer, not by stopping the chain.
Per-Step Agent Map (DOCTRINE rule 12)
| Step | Sub-phase | Worker tier | Thinking tier | Notes | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 — Resolve scope | — | — | — | Mechanical decision (exempt) | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 2 — Gather context | 2a — Surface mapping | Searcher × 2 (glob + import-graph) | Sonnet Reviewer | Parallel | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 2 — Gather context | 2b — Semantic indexing | Searcher × 2 (type-system + symbol-graph) | Sonnet Reviewer | Parallel | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 2 — Gather context | 2c — Convention scan | Searcher × 1 (test patterns + lint config) | Sonnet Reviewer | Justified single-angle | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 2 — Gather context | 2d — Aggregate coverage gate | — | Reviewer (Opus) verifies aggregate coverage | Thinking-tier coverage gate | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 3 — Review | 3a — L1+L2 (syntax/format/naming) | — | Reviewer (Opus) × 2 (different file groups) + Sonnet Reviewer aggregates verdicts | Parallel Opus pair; justified single-tier (Opus are the workers at L1-L2) | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 3 — Review | 3b — L3 (integration/security) | — | Reviewer (Opus) × 2 (integration + security) + Sonnet Reviewer aggregates verdicts | Parallel Opus pair; justified single-tier (L3 requires thinking-tier) | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 3 — Review | 3c — L4+L5 (perf/scale/a11y/UX) | — | Reviewer (Opus) × 2 (perf/scale + a11y/UX) + Sonnet Reviewer aggregates verdicts | Parallel Opus pair; justified single-tier (L4-L5 requires thinking-tier) | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 4 — Findings synthesis | 4a — Critical findings | Writer × 2 (evidence probe + impact analysis) | Sonnet Reviewer | Parallel |
| Subcommand | Description |
|---|---|
list |
Print the registry: in-flight · completed-uncollected · stalled · errored |
show |
Print one agent's output buffer (.hyperflow/background/) |
cancel |
Cancel one specific in-flight agent |
cancel --all |
Cancel every in-flight agent (use before closing a session) |
prune |
Delete completed .hyperflow/background/ files older than 7 days |
Default subcommand when none provided: list.
Subcommand Details
list
Read .hyperflow/background/registry.json. Group entries by status and print a compact table:
## In flight (N)
| ID | Purpose | Fired | Timeout | Blocks |
|-----------------------------------|--------------------------------------|------------|---------|---------|
| `bg-1718049600-quality-gates-b2` | Layer 5 gates Batch 2 | 17:30 | 18:00 | step3 |
| `bg-1718049820-ci-watcher` | GitHub Actions watch for v4.7.0 | 17:33 | 18:33 | — |
## Completed (uncollected, N)
| ID | Purpose | Completed | Duration | Output |
|-----------------------------------|--------------------------------------|------------|----------|--------|
| `bg-1718045400-scaffold-refresh` | Refresh .hyperflow/architecture.md | 16:42 | 2m 18s | 1.4kb |
## Stalled / Errored (N)
| ID | Purpose | Status | Reason |
|-----------------------------------|--------------------------------------|-------------------|-------------------|
| `bg-1717980000-cache-compact` | Compact learnings.md | STALLED | timeout (30m) |
Print one trailing line: . If registry is empty, print No background agents. and stop.
show
Read .hyperflow/background/ and print it verbatim. If the agent is still running, print the registry entry first then Output buffer not yet written.
Use when the user wants hyperflow's behavioral rules to apply outside the terminal CLI — in Claude Code Desktop, claude.
Bridge
Embed the portable subset of hyperflow's doctrine into the project's CLAUDE.md so it applies in surfaces that don't load CLI plugins (Claude Code Desktop, claude.ai web, IDE extensions). The doctrine block is managed via fenced markers, so refreshing it on plugin updates is idempotent and never touches your own CLAUDE.md content.
Source template: templates/claude-md-doctrine.md. Doctrine background: DOCTRINE.md.
Subcommands
| Subcommand | Description | ||
|---|---|---|---|
generate |
Write the doctrine block into the project's CLAUDE.md (create the file if absent, append the block if not present, refresh if already present) |
||
refresh |
Same as generate — alias for clarity when the block already exists |
||
remove |
Remove the doctrine block from CLAUDE.md (preserves your own content; if the file becomes empty after removal, leave it as an empty file rather than delete) |
||
status |
Show whether the doctrine block is present, its version, when it was generated | ||
`mode | manual\ |
off>` |
Set the auto-bridge mode for this project. Writes |
.hyperflow/.bridge-mode. The session-start hook reads this and decides what to do |
Default subcommand when none provided: status.
Auto-bridge (default ON)
The CLI session-start hook (hooks/session-start) runs scripts/auto-bridge.py on every session start. Behavior depends on the mode stored in .hyperflow/.bridge-mode:
| Mode | Behavior |
|---|---|
auto (default when .bridge-mode is absent) |
If ./CLAUDE.md is missing the doctrine block OR has an outdated version, silently writes/refreshes the block and prints a one-line notice in session-start output. Zero user friction. |
manual |
Never writes. Prints a one-line advisory when the block is missing or outdated: ./CLAUDE.md doctrine block would be refreshed (version 4.11.0) — run /hyperflow:bridge refresh to apply. |
off |
Does nothing. No writes, no advisories. |
This means: open Claude Code CLI once in your project, and from then on every Desktop / web / IDE session in the same project automatically gets the up-to-date hyperflow doctrine via CLAUDE.md. Refresh on plugin update is automatic too.
To opt out: /hyperflow:bridge mode off. To require explicit refresh
Use when the user wants to view, search, add, edit, prune, archive, or clear hyperflow memory entries.
Cache
CRUD interface for .hyperflow/memory/. Full protocol: memory-system.md.
Storage
All operations target .hyperflow/memory/ at the project root. Never modify source code files — if asked to "remember X about file Y", add a memory entry only, never edit Y.
Subcommands
| Subcommand | Description |
|---|---|
show [tag] |
Print index or filter entries by tag |
search |
Full-text search across all memory files |
add |
Append a new entry (prompts for details) |
edit |
Find entry by date+title slug and update in place |
prune |
Remove stale, superseded, and orphaned entries |
archive |
Move entries older than 30 days to cold storage |
clear |
Wipe all memory (with confirmation, recoverable) |
stats |
Counts, tier breakdown, tag frequency, oldest/newest |
migrate |
Import entries from legacy ~/.claude/hyperflow-memory.md |
off |
Disable memory writes for this session |
compact |
Summarise aged memory entries into stubs + monthly archive sidecars |
Subcommand Details
show [tag]
No arg → print index.md. With tag → filter all files for matching entries.
Output table: Date | Title | Tags | File | Tier
search
grep/ripgrep across learnings.md, decisions.md, pitfalls.md, patterns.md, conventions.md.
Return file:line + snippet, ranked by relevance.
add
Categories: learning decision pitfall pattern convention
Prompt via AskUserQuestion for: what, why it matters, tags (controlled vocab).
Append to the matching file using:
### [YYYY-MM-DD] <title> `[tag1, tag2]`
**What:** ...
**Why it matters:** ...
**Evidence:** ...
Update index.md with the new row.
edit
Locate by date+title slug. Show current value, prompt for new value, update in place.
prune
Per memory-system.md pruning protocol:
- Remove
[SUPERSEDED]entries older than 7 days - Remove entries whose referenced files no longer exist (
test
Use when ready to ship — runs pre-push gates (lint, typecheck, build, tests, security sweep), commits, releases, and pushes.
Deploy
No gate skipped, no failure ignored. If any gate fails, halt and report. Never --no-verify. Never bypass.
Failure recovery (rule 14). Worker errors and Quality Gate failures follow the canonical policy in skills/hyperflow/failure-recovery.md. Gate failures are user-surfaced, never auto-fixed — print the failing command + full stderr and halt the push. Never --no-verify, never force-push to main.
Per-Step Agent Map
| Step | Sub-phase | Worker tier | Thinking tier | Notes |
|---|---|---|---|---|
| 1a | Repo-state scan | Worker A (git status), Worker B (git log) | Sonnet | — |
| 1b | Tool detection | Worker A (profile.md + lockfiles), Worker B (testing.md + devDeps) | Sonnet | — |
| 2a | Lint + typecheck (parallel) | Worker A (linter), Worker B (formatter), Worker C (tsc) | Sonnet | Step 3 (Security Sweep) runs in parallel with Step 2 at orchestrator level; 2a halts chain on any failure before 2b |
| 2b | Build gate | Worker A (prod build), Worker B (dev build) | Sonnet | Depends on 2a PASS |
| 2c | Test gate | Worker A (unit), Worker B (integration/E2E) | Sonnet | Parallel (P1); depends on 2b PASS |
| 3a | Secrets scan | Worker A (diff pattern), Worker B (file pattern) | Opus | Runs in parallel with Step 2 (pre-build; read-only) |
| 3b | Dependency audit | Worker A (CVE audit), Worker B (license check) | Sonnet | — |
| 4 | Commit | single Worker | Sonnet | atomic-exempt (DOCTRINE 12.2) |
| 5a | Release execution | single Worker | Sonnet | atomic-exempt (DOCTRINE 12.2) |
| 5b | Version sync | Worker A (manifests), Worker B (changelog) | Sonnet | — |
| 6 | Push gate | AskUserQuestion | — | structural gate; atomic-exempt |
| 7 | Output | single print | — | atomic-exempt (§12.1) |
Step 1 — Survey State
Sub-phases run in parallel (P1).
Step 1a — Repo-state scan
Two Workers in parallel:
- Worker A —
git status --short— uncommitted changes, staged files - Worker B —
git log origin/— commits ahead of remote; detect branch name..HEAD --oneline
Sonnet Reviewer — verdict on repo state (clean / has uncommitted / ahead by N). If detached HEAD or no remote configured → halt with reason.
Step 1b — Tool detection
Use when a task file exists in .
Dispatch
Workhorse phase. Picks up a task file from /hyperflow:scope and runs it through the orchestrator pattern with parallel worker dispatch and thinking-tier reviews.
This skill exercises Layer 3 (Orchestrator), Layer 5 (Quality Gates), Layer 6 (Project Memory), Layer 8 (Git Workflow), and Layer 9 (Security) from the doctrine. Multi-level review (L1–L5) is applied per the triage's flow profile.
Per-Step Agent Map (DOCTRINE rule 12 — §12.1 inline-allowed for trivial steps · §12.2 sub-phase decomposition)
Every substantive step dispatches at least one Agent. Trivial steps (≤ 2 tool calls, no content generation, no decision-making, mechanically verifiable) MAY be performed inline by the orchestrator per §12.1. Non-trivial steps decompose into ≥ 2 named sub-phases per §12.2.
| Step | Sub-phase | Worker tier | Thinking tier | Notes | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 — Mode confirm | — (exempt) | — | — | AskUserQuestion only |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 0.5 — Operational choices | — (exempt) | — | — | AskUserQuestion only |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 1 — Load task | — (atomic · §12.2.8) | — | — | Read + schema check = single mechanical decision; no parallel angles | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 2a — Pre-dispatch | Composer × N parallel (Sonnet) — one per sub-task; stitches persona + injects learnings | Reviewer (Sonnet) — reviews prompt set for completeness | Parallel worker prompts built before any fan-out fires | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 2b — Worker fan-out | Implementer / Searcher / Writer × N parallel (Sonnet) | Reviewer (Sonnet · worker tier) — batched over full batch (P2) or per-sub-task fallback (mixed caps / --thorough) |
One Reviewer call per batch · escalates to Opus under --thorough |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 2c — Gate run | Worker (Sonnet) — runs lint/typecheck/tests on affected files | Reviewer (Sonnet) — judges gate output | Small focused diff; Sonnet sufficient | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 2d — Learnings + commit | Writer (Sonnet) — synthesizes per-batch learnings | — (mechanical commit · §12.1) | Per-sub-task PASS commits land here; learnings appended to context | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 3 — Final integration review | — (atomic · §12.2.8) | Reviewer (Opus · thinking tier) — L1–L |
Single Reviewer dispatch, no parallel angles; atomic-exempt | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 4 — Wrap up | Writer (Sonnet) — optional; only if memory prose is non-trivial | — | §12.1 trivial-inline; no Reviewer (D5) | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 5 — End of chain | — (exempt) | — | ONE Use when the user wants to manually flush a deferred-commit queue from a prior or interrupted chain.
ReadBash(git:*)Bash(ls:*)Bash(cat:*)Bash(rm:*)
FlushManually flush the deferred-commit queue from a chain that ran with Subcommands
What gets flushed
What happens if fast-forward isn't possibleIf the user's branch diverged from staging (e.g. the user committed manually mid-chain on the same branch),
The staging branch is preserved for the user to act on manually. The queue manifest stays in place so a future Flow
Overview
Prerequisites
"Use when applying Hyperflow's orchestration doctrine in Codex, Antigravity, or another single-agent surface. Hyperflow Doctrine (single-agent port)Apply Hyperflow's behavioral floor in surfaces that load skills but do not provide the full Claude Code multi-agent runtime. Runtime AdaptationCodex and Antigravity run one foreground agent. Where the full doctrine says to dispatch parallel workers under reviewers:
Codex Function RouterCodex loads Hyperflow as skills, not as native Claude-style slash commands. Treat these user messages as function aliases and execute the matching skill workflow inline in the current thread:
Do not answer that Hyperflow prompt amplifier. hyperflow-amplify — prompt amplifier (Antigravity single-agent)Turn a rough prompt into the single strongest version, then hand it off. Amplify never writes code — it produces a prompt other workflows (or you) execute. Follow the Steps
Rules
Hyperflow code review. hyperflow-audit — review phase (Antigravity single-agent)Multi-level review over a target (default: Levels
Default to L2; elevate to L3 when the diff touches auth, data, money, or external input. Steps
Rules
Hyperflow memory manager. hyperflow-cache — memory CRUD (Antigravity single-agent)Manage Operations
Rules
Hyperflow ship phase. hyperflow-deploy — ship phase (Antigravity single-agent)Gate, then ship. Follow the Steps
Hard rules
Hyperflow execution phase. hyperflow-dispatch — execution phase (Antigravity single-agent)Execute the task file from Per batch
After all batches
Rules
Hyperflow project setup. hyperflow-scaffold — project setup (Antigravity single-agent)One-shot setup of hyperflow's project surfaces. Follow the Steps
Rules
Hyperflow planning phase. hyperflow-scope — decomposition phase (Antigravity single-agent)Decompose, don't build. The only writes are to Steps
Rules
Hyperflow design phase. hyperflow-spec — design phase (Antigravity single-agent)This phase is thinking, not building. No code until the design is approved. Follow the Steps
Rules
Hyperflow project status. hyperflow-status — read-only state (Antigravity single-agent)One-screen view of hyperflow project state. Read-only — never edits files or runs work. Follow the Steps
Rules
Hyperflow auto-routing mode. hyperflow-sticky — auto-routing mode (Antigravity single-agent)Control hyperflow's auto-routing aggressiveness. Follow the Modes
Steps
Rules
Hyperflow debugging. hyperflow-trace — root-cause phase (Antigravity single-agent)Find the root cause before changing anything. Follow the Steps
Rules
Use when starting hyperflow in a new project, refreshing the .
ReadWriteEditBash(git:*)GlobGrep
ScaffoldOne-shot project setup. Analyzes the codebase, builds the Step 1 — Analysis CacheCheck for If absent — dispatch parallel searchers (single message, six Agent calls):
See project-analysis.md for what each file captures. If present — staleness check: Compute SHA256 of tracked config files, compare against After analysis:
Step 2 — Memory SkeletonCreate
session-context.md — populated by the session-start hook, not scaffold: Scaffold creates the empty Use when the user has a clear-enough task and wants it decomposed into batched worker sub-tasks before any code is written.
ReadWriteEditBash(git:*)GlobGrepAskUserQuestion
ScopeDecompose, don't build. Read-only with respect to source code. The only writes are to This skill exercises Layer 0 (Project Analysis) for context, Layer 6 (Project Memory) for past-learning surfacing, and Layer 7 (Task Templates) for decomposition patterns. It also inherits the triage classification from Iron Rules
Per-Step Agent Map (DOCTRINE rule 12 + 12.2)Every substantive step dispatches at least one Agent per DOCTRINE rule 12. Trivial steps per §12.1 may be performed inline by the orchestrator. Non-trivial Steps decompose into ≥ 2 named sub-phases per DOCTRINE rule 12.2.
|