app-audit-lite
Runs a quick App Store / Google Play audit of one app with live AppGoat data - listing quality, ratings trend, recent complaints, keyword rankings, competitors and their recent changes - and ends with a prioritized fix list.
App audit (lite)
A 10-minute audit of one app from public store data: what the listing says, how users rate it, what they complain about, which searches it wins, and what competitors changed recently. It ends with at most 7 fixes, ordered by expected impact.
This is the cheap "lite" path (~25–40 credits), assembled from the AppGoat read tools. The full
App Audit is live as the audit_app MCP tool (POST /v1/apps/{app}/audit): an AI-written,
evidence-checked report with a 0–100 score, prioritised actions, metadata rewrites within the
character limits, ~20 probed search terms and a shareable report page. It takes 1–3 minutes:
audit_app waits about 25 s and returns an audit id, which get_audit follows up on (free). The
first audit of an organization is free, later ones cost 100 credits, and re-running the same app,
country and mode within 24 h returns the earlier audit for free.
When to use
- The user wants an assessment of their own app or a competitor's ("audit", "teardown", "what should I fix").
- Before an ASO rewrite (run
aso-keywords-metadata-rewriteafterwards for the metadata itself). - Prefer
audit_appinstead when the user wants the full report or a link to share, or has not used their free audit yet (call it withfree_only: trueto run it only if it is free). Stay on this lite flow when credits are tight or for a quick check. - Not for idea hunting (use
find-app-ideas) or for turning reviews into tickets (usereviews-to-github-issues).
Inputs
app(required): anything that identifies the app - store URL, Apple id (284882215), Play package (com.spotify.music), bundle id or a name. If a name resolves to the wrong app (check the warning in the tool output), ask the user for the store link.country(optional, defaultus): two-letter storefront code. Audit one country per run.
Steps
Call the AppGoat MCP tools in this order. Costs are AppGoat credits; every tool output ends with a
footer showing credits_used and credits_remaining - stop and tell the user if remaining credits
drop below 50. Without the MCP, use the REST equivalents in reference/api.md.
| # | Call | Credits | What you need from it |
|---|---|---|---|
| 1 | get_app {app, country, include: ["versions", "charts"]} | 1 (+1 if older than 7 days and refreshed) | title, subtitle, description, category, rating, ratings_count, histogram, version, updated_at, release notes of the last 5 versions, screenshots count, languages, chart positions, linked app in the other store |
| 2 | get_app_history {app, country} | 1 | rating and ratings_count over the last 90 days |
| 3 | get_reviews {app, country, rating: "1-2", limit: 100} | 1 (+1 refresh) | recent complaints; group by theme yourself |
| 4 | get_reviews {app, country, rating: "5", limit: 50} | 1 | what users love (words to reuse in the listing) |
| 5 | get_app_keywords {app, country, limit: 100} | 1 | ranked terms, rank, popularity, difficulty |
| 6 | get_competitors {app, country, limit: 5} | 1 | top 5 competitors |
| 7 | get_app {app: , store, country} for each of the 5 | 5 | competitor title, subtitle, rating, ratings_count, updated_at |
| 8 | get_app_changes {app: , store, field: "title,subtitle,description,screenshots,price,version", since: <today − 60 days>} for the top 3 | 3 | what competitors changed recently |
| 9 | keyword_research {terms: up to 10 high-popularity terms competitors rank for but the app does not (from step 5 vs competitor titles/subtitles), store, country, app} | 10 | popularity, difficulty, the app's rank and relevance per gap term |
Budget: about 25 credits (up to ~40 with refreshes and a second reviews page). Do not page through more reviews unless the user asks.
If a call answers that data is still being refreshed (refreshing, a job_id), call get_job
{id} (free) every few seconds until it is done, then repeat the original call once.
Decision rules
Score each area Good / Needs work / Problem using these thresholds:
- Rating: Problem if rating < 4.0 or the 90-day history shows a drop ≥ 0.1; Needs work if 4.0–4.4; Good if ≥ 4.5. Flag if 1★ share of the histogram > 15%.
- Rating volume: Needs work if ratings_count grew < 1% in 90 days while competitors' median ratings_count is ≥ 5× higher (prompt for ratings in the app).
- Freshness: Problem if updated_at > 180 days ago; Needs work if > 60 days. Release notes that are generic ("bug fixes and improvements") in ≥ 3 of the last 5 versions: Needs work.
- Title / subtitle (Apple 30 / 30 characters; Play title 30, short description 80): Needs work if either uses < 70% of its limit, if the title has no non-brand keyword, or if the subtitle repeats title words.
- Keywords: count ranked terms with rank ≤ 10 and popularity ≥ 30 ("winning"), and terms with rank 11–50 and popularity ≥ 30 ("striking distance"). Problem if winning = 0; every striking-distance term with relevance < 0.7 is a metadata opportunity.
- Complaints: a theme counts when ≥ 3 of the 1–2★ reviews mention it, or ≥ 1 mentions a crash,
data loss, a billing/subscription problem or a login lockout. Unanswered 1–2★ reviews (no
reply)50%: Needs work.
- Competitors: note every competitor that changed its title or subtitle in the last 60 days (they are targeting new terms) and any competitor with rating ≥ 0.3 higher than the app.
- Localization: Needs work if
languageshas 1 entry and any competitor has ≥ 5.
Fix list: at most 7 items, each tied to a finding above, ordered by (impact × ease). Crashes and billing complaints always come first, then title/subtitle keyword gaps, then everything else. Never invent numbers; if a value is missing, say "no data".
Output format
# App audit: <title> (<store>, <country>) - <date>
**Verdict:** <one sentence>
| Area | Status | Evidence |
|---|---|---|
| Rating | Good / Needs work / Problem | 4.6★ (42,310 ratings), −0.05 in 90 days, 1★ 6% |
| Freshness | … | v3.2.1, updated 12 days ago |
| Title & subtitle | … | "…" (24/30), "…" (18/30) |
| Keywords | … | 4 winning, 9 striking distance |
| Complaints | … | login loop (7), sync lost data (2) |
| Competitors | … | X changed subtitle to "…" on <date> |
| Localization | … | English only; 3 of 5 competitors ≥ 5 languages |
## Top fixes
1. **<fix>** - why (evidence), expected effect.
2. …
## Keyword opportunities
| Term | Popularity | Difficulty | Your rank | Relevance | Action |
|---|---|---|---|---|---|
## What users say
- Love: "<≤ 120-char quote>" (5★, v3.2)
- Hate: <theme> - "<quote>" (1★, v3.2)
## Competitors
| App | Rating | Ratings | Updated | Recent change |
|---|---|---|---|---|
_Credits used: <sum of the footers>._
**Automate this →** weekly competitor changes and review alerts: see the `reviews-to-github-issues`
playbook; metadata fixes: `aso-keywords-metadata-rewrite`.
Automate it
Run the audit monthly and commit the report, so changes are visible in git history.
GitHub Action (.github/workflows/appgoat-audit.yml, secrets ANTHROPIC_API_KEY and APPGOAT_API_KEY):
name: AppGoat monthly audit
on:
schedule:
- cron: "0 7 1 * *"
workflow_dispatch:
permissions:
contents: write
jobs:
audit:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- name: Install AppGoat playbooks and MCP config
env:
APPGOAT_API_KEY: ${{ secrets.APPGOAT_API_KEY }}
run: |
git clone --depth 1 https://github.com/appgoat/skills /tmp/appgoat-skills
mkdir -p .claude/skills
cp -r /tmp/appgoat-skills/skills/* .claude/skills/
# Hosted AppGoat MCP server; the key goes into a file, not the command line.
jq -n --arg key "$APPGOAT_API_KEY" \
'{mcpServers: {appgoat: {type: "http", url: "https://api.appgoat.ai/mcp", headers: {"x-api-key": $key}}}}' \
> "$RUNNER_TEMP/appgoat-mcp.json"
- uses: anthropics/claude-code-action@v1
with:
anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
prompt: |
Use the app-audit-lite skill for the app https://apps.apple.com/us/app/id000000000 in country us.
Write the report to reports/audit-<YYYY-MM>.md (current month), then commit and push it.
claude_args: |
--mcp-config ${{ runner.temp }}/appgoat-mcp.json
--allowedTools mcp__appgoat__get_app,mcp__appgoat__get_app_history,mcp__appgoat__get_reviews,mcp__appgoat__get_app_keywords,mcp__appgoat__get_competitors,mcp__appgoat__get_app_changes,mcp__appgoat__keyword_research,mcp__appgoat__get_job,Write,Bash(git:*)
Claude Code routine (paste into Claude Code with the AppGoat connector added):
/schedule monthly on the 1st at 08:00 - Run the app-audit-lite playbook for <app link> in <country>. Compare it with last month's report in reports/, call out what improved and what got worse, and open a PR with the new report.
Related skills
- Downloads
- 67All time
- Security audit
- Review
- Last updated
- 4d ago
- Current version
- v1.0.0
- License
- MIT-0
