Pp Flow

<!-- GENERATED FILE — DO NOT EDIT.
     This file is a verbatim mirror of library/media-and-entertainment/flow/SKILL.md,
     regenerated post-merge by tools/generate-skills/. Hand-edits here are
     silently overwritten on the next regen. Edit the library/ source instead.
     See the repository agent guide, section "Generated artifacts: registry.json, cli-skills/". -->

Google Flow — Printing Press CLI

Prerequisites: Install the CLI

This skill drives the flow-pp-cli binary. You must verify the CLI is installed before invoking any command from this skill. If it is missing, install it first:

1. Install via the Printing Press installer. It defaults binaries to $HOME/.local/bin on macOS/Linux and %LOCALAPPDATA%\Programs\PrintingPress\bin on Windows:

   npx -y @mvanhorn/printing-press-library install flow --cli-only

2. Verify: flow-pp-cli --version
3. Ensure the reported install directory is on $PATH for the agent/runtime that will invoke this skill.

If the npx install fails (no Node, offline, etc.), fall back to a direct Go install (requires Go 1.26.6 or newer). This installs into $GOPATH/bin (default $HOME/go/bin), so add that directory to $PATH instead:

go install github.com/mvanhorn/printing-press-library/library/media-and-entertainment/flow/cmd/flow-pp-cli@latest

If --version reports "command not found" after install, the runtime cannot see the binary directory on $PATH. Do not proceed with skill commands until verification succeeds.

Flow has no in-app Google Drive picker and no way to import or sync to your own audio track -- this CLI pulls seed images straight from Drive, turns a Scribe recap script into ready-to-approve per-shot prompts, and muxes your real audio back onto the rendered clips locally, while staying honest that the final credit-spend click stays a transparent, user-driven action because of Google's reCAPTCHA gate on that step. The full pipeline is five steps, two of them manual by necessity: (1) episode import drafts the prompt queue from your two Drive folders, (2) you submit each shot's prompt in the real Flow UI (this CLI cannot automate that click -- see Authentication and Troubleshooting below), (3) you copy each returned job/workflow name into the queue file's job_name field, (4) video watch --batch polls them all at once, (5) mux lays your real audio back over the finished clips.

When to Use This CLI

Use this CLI for any Flow workflow that benefits from offline search over your own project library, scriptable batch preparation, Google Drive asset ingestion, or turning a timed audio-drama script into a ready-to-approve Flow prompt queue with a local audio mux-back step at the end.

Anti-triggers

Do not use this CLI for:

• Do not use this CLI to bypass or automate Google's reCAPTCHA-gated generation-submit step -- it is designed to stop short of that and hand off to your real browser.
• Do not use this CLI's TTS/character-voice needs -- Flow's Voices tab has no CLI surface here since it does not serve the audio-drama (real-mp3) workflow this CLI targets.
• Do not use this CLI for the official, documented Gemini API / Veo model access (that is a separate, fully-documented Google product surface with its own SDKs) -- this CLI targets the Flow web product specifically.

Unique Capabilities

These capabilities aren't available in any other tool for this API.

Audio-drama pipeline

• script draft-prompts — Turn a Scribe recap-script JSON into a ready-to-approve queue of per-segment Flow prompts, each matched to the right seed image automatically.

  Reach for this whenever the user has a Scribe-produced recap script and wants shot-by-shot Flow prompts drafted without hand-typing each one.

  flow-pp-cli script draft-prompts recap_script.json --images-dir ./seed-images --out episode3-queue.json

• episode import — Pull a whole episode's assets from two separate Google Drive folders -- Scribe's session output and the images folder -- into a Flow project and draft the prompt queue, in one command.

  Reach for this to start a new episode end-to-end from the two Drive folders instead of running drive import and script draft-prompts separately.

  flow-pp-cli episode import --scribe-folder ~/gdrive/session12-scribe --images-folder ~/gdrive/episode12-images

• mux — Lay your real audio-drama mp3 back over the rendered Flow clips, in the right order and at the right offsets, with one local command.

  Reach for this as the last step of any audio-drama animation pass -- it replaces the manual video-editor re-assembly step entirely.

  flow-pp-cli mux shot1.mp4 shot2.mp4 shot3.mp4 --audio episode3.mp3 --beats episode3-beats.json --out episode3-final.mp4

Batch economics

• queue estimate — See whether a prepared batch of generations fits your remaining Flow credits before you spend a single one.

  Reach for this before submitting any batch, especially when running multiple client projects against a shared credit pool.

  flow-pp-cli queue estimate episode3-queue.json

• video watch — Check on an entire submitted batch of generations with one command instead of clicking through each one.

  Reach for this after queuing several generations and walking away -- one glance at aggregate progress instead of N manual look-ups.

  flow-pp-cli video watch --batch episode3-queue.json

Drive ingestion

• drive import — Pull seed images straight out of a local Google Drive folder into a Flow project, with character tags inferred automatically from filenames (requires --project for character names).

  Reach for this whenever new reference images land in a Drive folder and need staging into a Flow project -- eliminates the download-then-reupload dance entirely.

  flow-pp-cli drive import --folder-id ~/gdrive/episode3-images --tag-scene --project a1b2c3d4-e5f6-47a8-9b0c-1d2e3f4a5b6c

Local sanity checks

• scenes gaps — Find characters missing a reference image and see a media-status breakdown for a project, before you submit a batch.

  Reach for this as a pre-flight check before a big batch submission to catch missing assets early.

  flow-pp-cli scenes gaps --project a1b2c3d4-e5f6-47a8-9b0c-1d2e3f4a5b6c

HTTP Transport

This CLI uses Chrome-compatible HTTP transport for browser-facing endpoints. It does not require a resident browser process for normal API calls.

Discovery Signals

This CLI was generated with browser-observed traffic context.

• Capture coverage: 0 API entries from 0 total network entries

Command Reference

credits — Flow/Veo credit balance

• flow-pp-cli credits — Check remaining Flow/Veo credit balance

flowWorkflows — Generation workflow lifecycle state

• flow-pp-cli flow-workflows <workflowId> — Fetch a generation workflow's current state

project — Full contents of one Flow project (media, characters/scenes, workflows)

• flow-pp-cli project — Fetch a project's full contents: media assets, character/scene entities

projects — Flow projects (offline-mirrored)

• flow-pp-cli projects — List the authenticated user's Flow projects

video — Async video generation jobs

• flow-pp-cli video — Poll status for one or more in-flight/queued video generations

Finding the right command

When you know what you want to do but not which command does it, ask the CLI directly:

flow-pp-cli which "<capability in your own words>"

which resolves a natural-language capability query to the best matching command from this CLI's curated feature index. Exit code 0 means at least one match; exit code 2 means no confident match — fall back to --help or use a narrower query.

Recipes

Stage a whole episode from Drive

flow-pp-cli drive import --folder-id ~/gdrive/episode3-images --tag-scene --project a1b2c3d4-e5f6-47a8-9b0c-1d2e3f4a5b6c

Pulls every new reference image out of a local Drive folder into the active project in one call, tagged with matching character names.

Draft a shot-by-shot queue from a recap script

flow-pp-cli script draft-prompts recap_script.json --images-dir ./seed-images --out episode3-queue.json

Mechanically merges each recap-script element with its matching seed image into a Flow-ready prompt, no hand-typing required.

Check a batch fits your credits before spending anything

flow-pp-cli queue estimate episode3-queue.json --api-key AIzaSyDUMMY00000000000000000000000000

Sums the queue's expected Veo-tier cost against your live balance and flags what to trim (omit --api-key for a local-only total with no live comparison).

Track a whole batch instead of babysitting one clip at a time

flow-pp-cli video watch --batch episode3-queue.json

Aggregates repeated status polls across every job id in the queue into one progress table.

List projects with just the fields you need

flow-pp-cli projects --json --select id,title,modifiedTime

Pairs --json with --select to avoid burning context on Flow's full nested project payload. Use the projects command directly, not sync --resources projects -- the generic sync/export engine can't build the tRPC input envelope this endpoint requires yet (see Troubleshooting below).

Finish the loop: mux your real audio back onto the rendered clips

flow-pp-cli mux shot1.mp4 shot2.mp4 shot3.mp4 --audio episode3.mp3 --beats episode3-beats.json --out episode3-final.mp4

Overlays the user's real audio-drama track onto the ordered, rendered clips locally -- the step Flow itself cannot do.

Auth Setup

Flow has two independent auth surfaces, both rooted in the same Google sign-in but neither auto-refreshing. (1) aisandbox-pa.googleapis.com (credits, video status, generation) needs a harvested ya29.* Bearer token: labs.google/fx/tools/flow is a client-rendered page whose JavaScript mints it via Google Identity Services, with no server-side endpoint this CLI can call instead -- open the page in a logged-in browser, open its network tab, copy the Authorization: Bearer ya29.... header value off any aisandbox-pa.googleapis.com request, and export FLOW_SESSION_TOKEN=<that value>. (2) labs.google's own Next.js BFF (project, projects, scenes gaps, drive import --tag-scene) needs a NextAuth session cookie instead -- capture __Secure-next-auth.session-token from the same logged-in browser (e.g. via a Playwright storage_state() export or any cookie-export tool) and run flow-pp-cli auth login --cookies-file storage-state.json once. Re-capture and re-import/re-export either credential whenever a command reports the session has expired (~1 hour for the Bearer token).

Run flow-pp-cli doctor to verify setup.

Agent Mode

Add --agent to any command. Expands to: --json --compact --no-input --no-color.

• Pipeable — JSON on stdout, errors on stderr

• Filterable — --select keeps a subset of fields. Dotted paths descend into nested structures; arrays traverse element-wise. Critical for keeping context small on verbose APIs:

  flow-pp-cli credits --api-key your-token-here --agent --select remainingCredits,planTier

• Previewable — --dry-run shows the request without sending

• Offline-friendly — sync/search commands can use the local SQLite store when available

• Non-interactive — never prompts, every input is a flag

• Explicit confirmation — --agent does not imply --yes; pass --yes separately only after the target, arguments, and side effects are clear

• Explicit retries — use --idempotent only when an already-existing create should count as success

Response envelope

Commands that read from the local store or the API wrap output in a provenance envelope:

{
  "meta": {"source": "live" | "local", "synced_at": "...", "reason": "..."},
  "results": <data>
}

Parse .results for data and .meta.source to know whether it's live or local. A human-readable N results (live) summary is printed to stderr only when stdout is a terminal AND no machine-format flag (--json, --csv, --compact, --quiet, --plain, --select) is set — piped/agent consumers and explicit-format runs get pure JSON on stdout.

Paths and state

Agents should treat the CLI's path resolver as part of the runtime contract:

• Use --home <dir> for one invocation, or set FLOW_HOME=<dir> to relocate all four path kinds under one root.

• Use per-kind env vars only when a specific kind must diverge: FLOW_CONFIG_DIR, FLOW_DATA_DIR, FLOW_STATE_DIR, FLOW_CACHE_DIR.

• Resolution order is per-kind env var, --home, FLOW_HOME, XDG (XDG_CONFIG_HOME, XDG_DATA_HOME, XDG_STATE_HOME, XDG_CACHE_HOME), then platform defaults.

• config contains settings like config.toml and profiles. data contains credentials.toml, data.db, cookies, and auth sidecars. state contains persisted queries, jobs, and teach.log. cache contains regenerable HTTP/cache files.

• Stored secrets live in credentials.toml under the data dir. Existing legacy config.toml secrets are read for compatibility and leave config.toml on the first auth write.

• Run flow-pp-cli doctor --fail-on warn to surface path and credential-location warnings. agent-context exposes a schema v4 paths block for agents that need the resolved dirs.

• For MCP, pass relocation through the MCP host config. The MCP binary does not inherit CLI flags:

  {
    "mcpServers": {
      "flow": {
        "command": "flow-pp-mcp",
        "env": {
          "FLOW_HOME": "/srv/flow"
        }
      }
    }
  }

Fleet precedence: an inherited per-kind env var overrides an explicit --home for that kind. Use FLOW_HOME or per-kind vars as durable fleet levers, and use --home only for a single invocation. Relocation is not reversible by unsetting env vars; move files manually before clearing FLOW_HOME, or doctor will not find credentials left under the former root.

Automatic learning

This CLI ships a self-capturing learning loop. The CLI does its own bookkeeping: every invocation is journaled locally, a failed flag followed by a corrected retry auto-derives a flag_alias candidate, and a teach on a query family without a playbook auto-synthesizes a playbook_candidate from the session's journal. Your job is judgment only: recall first, act on surfaced candidates, teach the final answer, playbook amend when you observe a correction. You never record failures by hand.

Step 1: recall before any discovery

Before list/search/drill commands on a new user question, run:

flow-pp-cli recall "<user's question>" --agent

The response envelope:

{
  "query": "...",
  "normalized": "<normalized form>",
  "query_entities": ["..."],
  "found": true | false,
  "match_score": 0.0,
  "results": [
    { "resource_id": "...", "resource_type": "...", "venue": "...",
      "confidence": 2, "entity_match": "exact|partial|unknown",
      "source": "taught|preseed|pattern", "warnings": ["..."] }
  ],
  "mismatches": [ /* only when --debug-mismatches */ ],
  "warnings": [ /* top-level */ ],
  "candidates": [
    { "id": 12, "class": "flag_alias | playbook_candidate",
      "summary": "...", "sightings": 3, "last_seen": "...",
      "rationale": "...",
      "next_action": ["<trial command>", "flow-pp-cli learnings confirm 12"] }
  ],
  "playbook": {
    "query_family": "...",
    "playbook": {
      "steps": [ { "cmd": "<command with {slot} substitution>", "purpose": "..." } ],
      "entity_slots": ["$ENTITY"],
      "expected_tool_calls": 3
    },
    "slots_resolved": { "$ENTITY": { "token": "<live token>", "canonical": "<canonical>" } },
    "notes": "<workarounds + gotchas for this query family>"
  },
  "notes": "<duplicate surface for non-playbook callers>"
}

Empty-store short-circuit: if the store has no learnings, playbooks, or candidates yet (recall finds nothing and learnings list and learnings candidates are both empty), skip recall for the rest of this session instead of taxing every query; resume recall-first once something has been taught.

Step 2: decision tree

Read candidates, playbook, notes, results[0], and warnings in that order:

if Candidates present (warnings include "candidates_present"):
    -> candidates are try-then-confirm, never facts. Follow each candidate's
       two-step next_action verbatim: run the trial command first, then run
       `learnings confirm <id>` only after the trial verified the behavior.
       Reject a wrong candidate with `learnings reject <id>`.
    -> NEVER re-teach something recall surfaced as a candidate; confirm or
       reject that candidate instead of teaching a duplicate.
    -> candidates ride alongside playbooks and resource hits, not instead of
       them; continue with the branches below after acting on them.

if Playbook present:
    -> READ Playbook.notes verbatim FIRST (workarounds + gotchas the CLI surface doesn't expose)
    -> replay Playbook.steps in order, substituting Playbook.slots_resolved entries
       for the entity slot tokens. If a step's slot is unresolved, fall back to
       discovery for that step only.
    -> the Playbook's expected_tool_calls is a budget; if you find yourself running
       materially more, record the divergence via `flow-pp-cli playbook amend`
       at end-of-session.

elif Notes present (no Playbook):
    -> read Notes verbatim before any discovery step; they carry known gotchas
       for this query family even when no structured choreography exists yet.

elif Found AND Results[0].EntityMatch == "exact" AND Results[0].Confidence >= 2:
    -> skip discovery; fetch live data for Results[*].ResourceID in parallel

elif Found AND Results[0].EntityMatch == "partial":
    -> candidate hint, NOT a hit; read the resource title to validate before trusting

elif (any row in Mismatches[] when --debug-mismatches was passed):
    -> treat as cold start; the stored learning is for a different entity
       (different canonical resolved from query_entities)

else:  // Found == false, no playbook, no notes
    -> cold start; run discovery normally; teach the answer afterward (Step 4).
       If the family has no playbook yet, that teach auto-synthesizes a
       playbook candidate from this session's journal - you do not need to
       record one by hand.

Playbook and Notes are orthogonal to the per-resource path. A recall response can carry both a Playbook AND a Results[] hit - use both: the Playbook tells you which choreography to run; the resource hits short-circuit specific steps. Default to skipping mismatches; pass --debug-mismatches only when investigating cold-start surprises.

Candidate judgment details: learnings confirm <id> prints the candidate's full payload before materializing it - check that the printed payload matches the behavior you verified. learnings reject <id> tombstones the derivation signature so the same candidate does not resurface. The envelope carries only the few candidates worth acting on now; flow-pp-cli learnings candidates lists the full open set.

Graceful degradation: if learnings confirm is an unknown command, you are driving an older binary - ignore the candidates guidance and follow the rest of the protocol.