aso-pipeline

/aso-pipeline

Multi-agent ASO optimization pipeline. Composes 23 aso-skills into a structured pipeline that analyzes competitors, discovers keywords, crafts metadata, defines creative strategy, localizes for target markets, optimizes ratings, and sets up tracking — with cross-phase insights no single skill produces alone.

Architecture:

• This skill: entry point, preflight, mode routing, score mode (inline)
• aso-director agent: pipeline orchestrator (state, checkpoints, signals, phases)
• 8 specialist agents: one per pipeline phase (thin composition layers over aso-skills)

Mode Detection

Parse mode from $ARGUMENTS:

Argument	Mode	Action
(none) or run	Full pipeline	Spawn director in run mode
score	Instant scorecard	Run inline (no agent spawn)
express	Quick keywords + metadata	Spawn director in express mode
audit	ASO health check	Spawn director in audit mode
diff	What changed since last run	Spawn director in diff mode
build	Scaffold ASO project	Spawn director in build mode
status	Show progress	Read state file inline (no agent spawn)
competitive keywords metadata creative localization ratings tracking collision	Single phase	Spawn director in phase mode with phase name

Phase 0: Preflight Check

Before spawning the agent, validate ALL dependencies. Run ALL checks before reporting — never fail one-at-a-time.

Required Plugins

aso-skills:

ls ~/.claude/plugins/cache/aso-skills/ 2>/dev/null

Provides: competitor-analysis, competitor-tracking, market-pulse, market-movers, keyword-research, seasonal-aso, metadata-optimization, android-aso, screenshot-optimization, app-icon-optimization, ab-test-store-listing, in-app-events, app-clips, localization, rating-prompt-strategy, review-management, crash-analytics, app-analytics, asc-metrics, aso-audit, app-marketing-context, app-launch, app-store-featured

Optional Checks

• Exa MCP: Check if mcp__plugin_everything-claude-code_exa__web_search_exa is accessible. If absent, warn: live competitor/market data unavailable — LLM knowledge only.
• DataForSEO MCP: Check if mcp__plugin_everything-claude-code_dataforseo__* tools exist. If absent, warn: no verified keyword volumes — qualitative tiering mode only.
• docs/proposal.md: If exists, product context available for brief seeding.

On Failure

Preflight FAILED. Missing dependencies:

PLUGINS:
  - aso-skills — Install:
    /plugin marketplace add Eronred/aso-skills
    /plugin install aso-skills@aso-skills

OPTIONAL:
  - Exa MCP not detected — competitor research will use LLM knowledge only.
  - DataForSEO not detected — keyword data will be qualitative (Tier A/B/C), not numeric.

Install missing dependencies and run /aso-pipeline again.

STOP here. Do NOT proceed until required plugins are installed.

On Success

Preflight passed.
  Plugins: aso-skills ✓
  Optional: Exa [✓/missing], DataForSEO [✓/missing]
  Proposal: docs/proposal.md [found/not found]
  Prior state: aso/.pipeline-state.json [found/not found]

Score Mode (inline — no agent spawn)

If mode is score, run the scorecard inline without spawning the director. This is the fastest path to value (~5K tokens).

Requirements

Need EITHER a store URL to fetch listing data OR user-provided answers.

Process

1. Ask for store URL (App Store or Play Store) if not in existing app_brief.yaml
2. Attempt to fetch the public listing page via WebFetch or Exa
3. If fetch succeeds, extract: title, subtitle, description length, screenshot count, video presence, rating, review count, localization count
4. If fetch fails, ask the user to provide:
   • Current title and subtitle
   • Number of screenshots and whether video exists
   • Current rating and review count
   • Number of locales supported
5. Apply 10 heuristic checks:

Check	Scoring	Max Points
Title character utilization (30 chars)	% used	10
Subtitle utilization (iOS: 30 chars)	% used, 0 if empty	10
Feature keyword in title	present = 10, absent = 0	10
Screenshot count (min 6 iOS, 8 Android)	meets min = 10, below = proportional	10
App preview video	present = 10, absent = 5	10
Rating (target: 4.5+)	4.5+ = 10, 4.0-4.5 = 7, 3.5-4.0 = 4, <3.5 = 0	10
Review volume	>1000 = 10, 100-1000 = 7, 10-100 = 4, <10 = 0	10
Description length (target: >500 chars)	>500 = 10, 200-500 = 5, <200 = 2	10
Localization coverage	5+ locales = 10, 2-4 = 5, 1 = 0	10
Promotional text utilization (iOS)	non-empty = 10, empty = 0	10

6. Output scorecard with letter grades (A: 9-10, B: 7-8, C: 5-6, D: 3-4, F: 0-2)
7. Add Top priority line: the single most impactful fix
8. Offer upgrade: "Want me to fix these? /aso-pipeline express or /aso-pipeline run"

If data was user-provided (not fetched), label: "Score based on self-reported data."

Status Mode (inline — no agent spawn)

If mode is status, read the state file inline:

1. Check if aso/.pipeline-state.json exists
2. If not found: "No pipeline state found. Run /aso-pipeline build to start."
3. If found: read and display phase status table:

   ASO Pipeline Status — [app_name]
   Started: [date]  |  Mode: [mode]  |  Maturity: [level]
   
   | Phase | Status | Completed | Signals |
   |-------|--------|-----------|---------|
   | Setup | ✓ | Apr 4 | — |
   | Competitive | ✓ | Apr 4 | SATURATED_MARKET |
   | Keywords | in progress | — | — |
   | ... | pending | — | — |
   
   Next: Keywords phase is in progress.
   Run: /aso-pipeline keywords to continue.
   

Cost & Time Estimate

After preflight passes (for modes that spawn the director), show estimate:

Mode	Tokens	Maturity Adjustment
run (established)	~247K	Full 8 phases
run (pre_launch)	~140K	Phases 0-4 only
run (early)	~180K	Phases 0-4 + 6
express	~58K	Phases 0, 2, 3
audit	~35K	Audit skill + recommendations
diff	~15K	Snapshot comparison
build	~12K	Interactive brief only
Single phase	~20-40K	Varies by phase

Spawn the Director

After preflight passes and estimate is shown:

Agent(
  subagent_type: "nextc-aso:aso-director",
  model: "sonnet",
  prompt: """
  Mode: [run/express/audit/diff/build/phase-name]
  Phase (if single-phase): [phase name]
  Working directory: [cwd]
  Exa available: [yes/no]
  DataForSEO available: [yes/no]
  Proposal exists: [yes/no]
  Prior state exists: [yes/no]
  Prior snapshots exist: [yes/no]

  Data quality: [ESTIMATED/MIXED/VERIFIED]
  - ESTIMATED: no DataForSEO, no AppTweak, no ASA → qualitative keyword tiering
  - MIXED: some verified sources + LLM estimates
  - VERIFIED: DataForSEO or AppTweak connected → numeric keyword scoring
  """
)

Mode-Specific Preconditions

Mode	Precondition	On Failure
run	None	Proceed (brief built interactively)
express	None	Proceed
audit	None	Proceed (brief built if needed)
diff	aso/.snapshots/ must have entries	"No snapshots found. Run the pipeline first, then use diff to track changes."
build	None	Proceed
status	aso/.pipeline-state.json must exist	"No pipeline state found."
Single phase	aso/config/app_brief.yaml must exist	"No app brief found. Run /aso-pipeline build first."
collision	3+ phases completed in state	"Collision needs 3+ completed phases."