rsa-maker

Create Google Ads Responsive Search Ads from offer angles and brand context, outputting a CSV ready for Editor import. Use to create, build, or generate RSAs from angles.

SKILL.md

GitHub

RSA Maker Skill

Generate Google Ads Responsive Search Ads from offer angles and brand context, outputting CSV files ready for import into Google Ads Editor.

Command Format

/rsa-maker [--new]

Examples:

  • /rsa-maker - Interactive mode, asks existing or new campaign
  • /rsa-maker --new - Skip to manual name entry

Process

Step 1: Prerequisites Check

Check required context files:

File Required If Missing
context/offer-angles.md Yes Prompt: "Run /offer-maker angles first"
context/brand.md Yes Prompt: "Run /ads-context [URL] first"

If both missing, stop and show:

## Prerequisites Missing

To create quality RSAs, the following context is required:

| File | Status | Action |
|------|--------|--------|
| context/offer-angles.md | Missing | Run `/offer-maker angles` first |
| context/brand.md | Missing | Run `/ads-context [URL]` first |

If one exists, show which is missing and the action to take.

Step 2: Campaign/Ad Group Selection

Question: "Is this for an existing campaign/ad group or a new one?"

Use AskUserQuestion tool:

  • Option 1: "Existing campaign/ad group" - Select from account data
  • Option 2: "New campaign/ad group" - Enter names manually

If existing (--new flag NOT provided):

  1. Read context/google-ads/data/campaigns.csv → extract unique campaign.name values
  2. Read context/google-ads/data/ads.csv → extract unique ad_group.name values
  3. Present campaign options via AskUserQuestion (max 4 options + Other)
  4. If more than 4 campaigns, show top 4 by impressions and allow "Other" for manual entry
  5. After campaign selection, show ad groups for that campaign

If new (or --new flag):

  1. Ask: "Enter the exact campaign name:"
  2. Ask: "Enter the exact ad group name:"

Step 2.5: Peer Context Pull (Mode 2 — Enrichment, no gate)

Before composition, read fresh peer audit reports and extract findings that should bias the headlines, descriptions, and CTAs you generate. This is purely read + extract. No gate, no warn-and-abort, no hard-block. If a peer report is missing or stale, skip that row silently — generation continues either way. The maker's output quality is what improves; peer findings flow into the generated asset.

2.5.1 Walk the 10-peer freshness table.

For each peer, check that the report file exists and is fresh per the window below. Freshness rule: the report header date is canonical. Read the first 30 lines of each peer file and locate the header date (typically **Generated:** YYYY-MM-DD or similar). If the file's mtime is older than the header date, surface a warning in the user-facing output (the file may have been edited but the header wasn't refreshed — treat the header as the source of truth, but flag the discrepancy). If the header date is within the fresh window, the peer is FRESH. Otherwise STALE — skip silently.

Peer skill Report file Fresh window Highest-leverage signal for RSAs
/quality-score-auditor context/analysis/quality-score-audit.md ≤ 7 days Ad Relevance, Expected CTR per AG
/search-term-auditor context/analysis/search-term-audit.md ≤ 7 days Top converting n-grams to mirror in headlines
/keyword-auditor context/analysis/keyword-audit.md ≤ 7 days Per-AG keyword themes for headline pinning
/budget-auditor context/analysis/budget-audit.md ≤ 7 days Which AGs are budget-limited (lower priority)
/bidding-auditor context/analysis/bidding-audit.md ≤ 7 days Smart-bidding context
/competitive-analyst context/analysis/competitive-audit.md ≤ 14 days Differentiation angles vs competitors
/lp-auditor context/analysis/lp-audit.md ≤ 14 days Message-match alignment to LP H1/offer
/offer-auditor context/analysis/offer-audit.md ≤ 30 days Offer angles to lead with
/tracking-specialist context/analysis/tracking-audit.md ≤ 30 days Informational warning if dirty
/strategy-specialist context/analysis/strategy-audit.md ≤ 30 days Posture/break-even informs CTA urgency

(/account-auditor is not part of the table — RSAs are about ad-group themes, not naming conventions. If context/analysis/account-audit.md is fresh and contains AG-level findings that look material, you may include it ad-hoc.)

2.5.2 For each FRESH peer, extract findings relevant to the selected campaign/ad group and quote them into a peer-context buffer.

Build an in-memory buffer keyed by peer skill. For each fresh peer, scan the report for the selected ad-group name (and selected campaign name, as fallback). Extract:

Peer What to extract
QS auditor Handoff Queue rows (Ad Relevance, Expected CTR) for the target AG. Note dominant_issue, recommended_action, class (BRANDED/COMPETITOR/PROBLEM/SOLUTION).
Search-term auditor Top converting n-grams for the target AG (from the n-gram module). 3-7 phrases the user actually searches and converts on.
Keyword auditor Per-AG keyword theme — the dominant intent cluster + top-volume keywords for this AG.
Budget auditor Is this AG's campaign budget-limited? (Yes/No flag — informs how aggressively to lean on USP vs price-anchor language.)
Bidding auditor Smart-bidding strategy + learning state for this campaign (informs whether the AG is in a stable optimization regime or still volatile).
Competitive analyst Differentiation angles vs top competitors for this campaign's keyword set.
LP auditor Message-match findings for the target Final URL — the LP's H1, primary offer phrasing, and any flagged mismatches.
Offer auditor Strongest offer angle (guarantee, value prop, urgency, trust) per offer-audit's recommendations for this campaign/AG.
Tracking specialist Module M (Measurement) status — clean or dirty?
Strategy specialist Account posture (growth/maintain/efficiency) and break-even constraints — informs CTA urgency.

Quote the actual finding text into the buffer (don't just note "QS audit fresh" — pull the row). Example buffer entry:

QS auditor (FRESH, 2 days old):
  - Ad group "AI Slides — Exact" has 4 keywords with AR Below Avg.
  - dominant_issue: "headline does not contain keyword phrase"
  - recommended_action: "Tighten H1 anchoring via Method B (Static Anchoring) — pin the
    keyword 'AI presentation maker' to H1 across all RSAs."
  - class: SOLUTION

Search-term auditor (FRESH, 1 day old):
  - Top converting n-grams for AG "AI Slides — Exact":
    1. "ai presentation maker" (28 conv)
    2. "make slides with ai" (19 conv)
    3. "ai powerpoint generator" (12 conv)
    4. "create slides ai" (8 conv)

2.5.3 Surface peer-report contradictions.

After building the buffer, scan for cross-peer contradictions and quote both sides. Do NOT auto-pick a winner. Examples to watch for:

  • LP auditor says LP H1 is "X" but offer auditor recommends leading offer angle "Y" (which doesn't appear on the LP).
  • QS auditor's recommended class (e.g., BRANDED) conflicts with the user's Step 3 traffic temperature (e.g., they classified Cold).
  • Search-term auditor's top n-gram differs from keyword auditor's dominant theme for the same AG.

If contradictions exist, surface them in the user-facing summary (Step 8) so the user picks which signal to align to. Quote both verbatim.

2.5.4 Tracking-audit informational note.

If tracking-audit.md is FRESH and Module M (Measurement) is dirty (any HIGH-severity findings, or the M-layer score is < 70), surface this one-liner once in the user-facing output (Step 8 summary header):

"Note: tracking-audit shows M layer dirty — RSAs will be created but performance measurement is unreliable until tracking is fixed."

Informational only. Does NOT block generation.

2.5.5 Header-vs-mtime warning.

If any peer report's filesystem mtime is older than its header date, surface this in the Step 8 summary:

"Warning: {peer-report.md} header date is {YYYY-MM-DD} but file mtime is older — header treated as canonical, but the report may have been edited inconsistently. Verify before relying on findings."

2.5.6 No gate.

If zero peer reports are fresh, generation still proceeds. The peer-context buffer is simply empty, and the Step 8 "Cross-skill context applied" section will say "No fresh peer audits available — RSAs composed from offer-angles.md and brand.md only." This is Mode 2 enrichment, not a Mode 1 gate.

Step 3: Traffic Temperature Classification

Different traffic types need different angle emphasis. Check if the selected campaign matches the classification in offer-angles.md.

3.1 Read existing classification:

Parse context/offer-angles.md → Traffic Classification section:

  • Campaign/Traffic name
  • Awareness Stage
  • Temperature (Cold/Warm/Hot)

3.2 Decision logic:

Scenario Action
Campaign name matches offer-angles classification Use existing temperature, confirm with user
Campaign name differs OR new campaign Ask user to classify traffic temperature

3.3 If classification needed, use AskUserQuestion:

Question: "What type of traffic will this RSA serve?"

Options:

Option Description
Brand Search (Hot) People searching for your brand name
Competitor Search (Hot) People searching for competitor names
Product/Service Search (Warm) People searching for what you sell (e.g., "AI presentation maker")
Problem/Symptom Search (Cold) People searching for their problem (e.g., "how to make slides faster")
Remarketing (Hot) Previous visitors returning

3.4 Set angle priorities based on temperature:

Angle Cold Warm Hot
Problem/Pain LEAD (H2-H3) Include (H8) Light
Value Proposition Include (H4) LEAD (H2) Include
USPs Light LEAD (H3, H6) LEAD
Value Boosters Skip Include LEAD
Social Proof Light Include (H4) LEAD
Risk Removal Skip Include (H5) LEAD

Priority meanings:

  • LEAD = Multiple headline slots (H2, H3, H6)
  • Include = At least one headline slot
  • Light = Optional, in descriptions only
  • Skip = Don't emphasize

Step 4: Gather Additional Inputs

Collect the following using AskUserQuestion:

  1. Final URL (required):

    • Question: "What is the Final URL? (e.g., https://plusai.com/)"
    • Validate: Must start with https://

  2. Path 1 (optional):

    • Question: "Display URL path 1 (max 15 chars, leave blank to skip):"
    • Validate: Max 15 characters

  3. Path 2 (optional):

    • Question: "Display URL path 2 (max 15 chars, leave blank to skip):"
    • Validate: Max 15 characters

  4. Pinning (optional):

    • Question: "Should specific headlines/descriptions be pinned to positions?"
    • Options:
      • "Yes - Pin key messages (Recommended)" - Claude selects up to 3 pins
      • "No - Let Google optimize all positions"

Step 5: RSA Composition

Read context/offer-angles.md and context/brand.md to compose RSA assets. Apply the peer-context buffer from Step 2.5 to bias the generated headlines, descriptions, and CTAs.

Reference files to read for this step:

  • references/rsa-composition-sop.md — Assembly framework, slot distribution, pinning rules
  • references/headline-angle-catalog.md — Headline patterns by type (7 types with formulas + examples)
  • references/description-expansion-catalog.md — Description patterns (single-angle + combined-angle)
  • references/headline-quality-checklist.md — Validate headlines after composition
  • references/description-quality-checklist.md — Validate descriptions after composition

Use the traffic temperature from Step 3 to determine slot distribution.

5.0 Peer-Context Biasing Rules

The peer-context buffer from Step 2.5 changes what gets generated (not whether). Apply the rules below in order — each one points to the slots it influences and the precise effect on copy. Track which rules fired so the Step 8 summary can list them under "Cross-skill context applied."

Trigger (peer finding) Effect on generation
Search-term auditor FRESH with top n-grams for target AG Mirror the top 3-5 n-grams verbatim (or near-verbatim, respecting 30-char limit) in headlines. Distribute across H2-H4 by search volume — top n-gram → H2, second → H3, third → H4. Do not stuff; one n-gram per headline max.
Keyword auditor FRESH with per-AG keyword theme Use the dominant theme keyword as the H1 anchor (overrides default {KeyWord:Default} framing — write a static H1 containing the theme keyword for stable Static Anchoring). Use 1-2 secondary cluster keywords in H6 / H8.
QS auditor FRESH + Ad Relevance Below Avg for target AG Prioritize tight keyword-mirroring in pinned headlines. Pin H1 to position 1 with the target AG's primary keyword phrase verbatim. Use Method B (Static Anchoring) — drop {KeyWord:Default} in favor of a hardcoded keyword headline. Avoid abstract benefit framing in H1-H3; benefits go to H4-H8.
QS auditor FRESH + Expected CTR Below Avg for target AG Lead H2-H4 with strong benefit / value-prop language and clear CTAs. Front-load specificity (numbers, outcomes, time savings). Description D1 must combine value-prop + clear next-step CTA. Avoid passive / generic copy.
QS auditor handoff class = BRANDED for target AG Branded low-QS is usually LP message-match — surface the BRANDED-class warning to the user (informational), and prioritize H1 = brand name + primary product, with descriptions mirroring the brand promise from brand.md.
QS auditor handoff class = COMPETITOR for target AG AR Below Avg is structurally expected for COMPETITOR campaigns — don't force tight keyword-mirroring. Lead with differentiation USPs and offer strength (per offer-audit) rather than relevance methods.
LP auditor FRESH + Message Match flagged for target Final URL Headlines and descriptions must align with the LP's H1 and primary offer copy. Quote the LP H1 into the buffer; H2 of the RSA should echo or paraphrase the LP H1 within 30 chars. D1 must include the LP's primary offer phrasing. If the LP H1 is materially different from offer-angles.md's value-prop, flag the contradiction (Step 2.5.3) and ask the user which to align to in Step 8.
Offer auditor FRESH with strongest offer angle RSAs lead with the strongest offer angle per offer-audit's recommendations. If "guarantee" is strongest → H4 or H5 = risk-removal headline + D2 = guarantee + CTA. If "value-prop" is strongest → H2 = value-prop. If "urgency" is strongest → H7 (CTA) carries urgency phrasing + D3 = urgency + risk-removal. Override the temperature-based slot defaults where offer-audit's recommendation is stronger.
Budget auditor FRESH + AG's campaign IS budget-limited Lower priority for this AG — keep RSAs simpler and tighter (fewer headlines, target 7 not 15). Lead with high-CTR / high-conversion-rate language to maximize value of each impression. Note in Step 8 summary: budget-limited campaign, RSAs designed for impression efficiency.
Budget auditor FRESH + AG's campaign NOT budget-limited Standard priority — full 15 headlines, full 4 descriptions, broader angle coverage.
Bidding auditor FRESH + smart bidding in LEARNING state RSAs should not introduce drastic copy variance during learning — use 1-2 dominant angles per RSA rather than 5+ competing angles. This stabilizes the bidding signal. Note in Step 8 summary.
Bidding auditor FRESH + tCPA / tROAS strategy Lean into proof / value-prop language that aligns with the conversion goal — tCPA campaigns favor lead-gen / signup CTAs; tROAS campaigns favor revenue / value-driven CTAs.
Competitive analyst FRESH with differentiation angles Use 1-2 differentiation angles in H3 / H6 (USP slots). Quote the differentiator verbatim where character limit allows. Avoid generic "best in class" framing — use the specific axis the competitive audit identified.
Strategy specialist FRESH + posture = growth CTA tone = inviting, low-friction ("Start free trial", "See it in action").
Strategy specialist FRESH + posture = efficiency or break-even tight CTA tone = decisive, high-intent ("Get a quote", "Book a demo"). Lean on risk-removal in D2.
Tracking specialist FRESH + M-layer dirty No effect on copy. Surface the informational note in Step 8 summary header (per Step 2.5.4).

Conflict resolution between peer rules:

  • When two rules target the same slot (e.g., search-term-auditor wants H2 = top n-gram, but offer-auditor wants H2 = value-prop), the closer-to-traffic peer wins: search-term > keyword > QS > offer > LP > competitive > bidding > budget > strategy > tracking. Note the conflict and the resolution in the Step 8 summary.
  • When peer rules conflict with the temperature-based default slot distribution (Step 3.4), peer rules win — they are based on observed account data, not heuristics.
  • When peer reports themselves contradict (e.g., LP H1 ≠ offer-audit's recommended angle), do NOT auto-pick. Surface both in Step 8 and ask the user.

Track which rules fired in a peer_rules_fired list — this drives the Step 8 "Cross-skill context applied" section. Each entry: {rule_name, slot_affected, source_quote}.

Slot Distribution by Temperature

Cold Traffic (Problem/Symptom Search):

Slot Angle Source
H1 Relevance Anchor {KeyWord:Default}
H2 Problem/Pain (LEAD) offer-angles.md → Problem/Pain phrases
H3 Problem/Pain (LEAD) offer-angles.md → Problem/Pain phrases (2nd)
H4 Value Proposition offer-angles.md → Value Prop phrases
H5 USP offer-angles.md → USP phrases
H6 Social Proof offer-angles.md → Social Proof phrases
H7 CTA brand.md → Primary CTA

Warm Traffic (Product/Service Search) - DEFAULT:

Slot Angle Source
H1 Relevance Anchor {KeyWord:Default}
H2 Value Proposition (LEAD) offer-angles.md → Value Prop phrases
H3 USP (LEAD) offer-angles.md → USP phrases
H4 Social Proof offer-angles.md → Social Proof phrases
H5 Risk Removal offer-angles.md → Risk Removal phrases
H6 USP (LEAD secondary) offer-angles.md → USP phrases (2nd)
H7 CTA brand.md → Primary CTA
H8 Problem/Pain (optional) offer-angles.md → Problem/Pain phrases

Hot Traffic (Brand/Competitor/Remarketing):

Slot Angle Source
H1 Relevance Anchor {KeyWord:Default}
H2 USP (LEAD) offer-angles.md → USP phrases
H3 Value Booster (LEAD) offer-angles.md → Value Booster phrases
H4 Social Proof (LEAD) offer-angles.md → Social Proof phrases
H5 Risk Removal (LEAD) offer-angles.md → Risk Removal phrases
H6 USP (LEAD secondary) offer-angles.md → USP phrases (2nd)
H7 CTA brand.md → Primary CTA
H8 Value Booster (optional) offer-angles.md → Value Booster phrases (2nd)

Headlines (7-8 total, max 30 chars each)

Composition guidelines:

  • Each headline MUST be ≤30 characters
  • Use exact phrases from offer-angles.md when possible
  • Use patterns from references/headline-angle-catalog.md to write headlines for each angle type
  • Adapt CTA from brand.md to fit character limit
  • If a phrase exceeds 30 chars, truncate or rewrite concisely
  • Include character count for verification
  • Follow the slot distribution for the selected traffic temperature

After composing all headlines, validate against references/headline-quality-checklist.md. Fix any failures before proceeding to descriptions.

Descriptions (2-3 total, max 90 chars each)

Slot Pattern Construction
D1 Problem + Solution Combine pain point + value prop (from offer-angles.md)
D2 Proof + CTA Social proof + action (from offer-angles.md + brand.md)
D3 (Optional) Offer + Risk Removal Value booster + guarantee (from offer-angles.md)

Composition guidelines:

  • Each description MUST be ≤90 characters, target 75-90
  • Use patterns from references/description-expansion-catalog.md (Section A for single-angle, Section B for combined-angle)
  • D1: Lead with problem, follow with solution (Catalog → Pattern 1: Problem + Solution)
  • D2: Lead with proof (numbers), end with CTA (Catalog → Pattern 2: Proof + Benefit)
  • D3: Include bonus value + risk removal if content available (Catalog → Pattern 3: Benefit + Risk Removal)

After composing all descriptions, validate against references/description-quality-checklist.md. Fix any failures before generating CSV.

  • Include character count for verification

Step 6: Pinning Logic (if enabled)

Apply up to 3 pins total. Selection criteria:

Asset Pin to Position Rationale
H1 (Relevance Anchor) Position 1 Ensures keyword visibility in first headline
D1 (Problem + Solution) Position 1 Strongest description appears first
H7 (CTA) Position 3 Ends with clear action

Position values for CSV:

  • 1, 2, 3 = Pinned to that position
  • - = Unpinned (Google optimizes)
  • Empty = Slot not used

If pinning disabled: All position columns get - for used slots, empty for unused.

Step 7: Generate CSV

Use the helper script to ensure proper CSV formatting.

7.1 Create JSON input file:

Create a JSON file at created/rsas/{timestamp}_input.json with the RSA data:

[
  {
    "campaign": "Campaign Name",
    "ad_group": "Ad Group Name",
    "headlines": [
      {"text": "Headline 1 text", "position": "1"},
      {"text": "Headline 2 text", "position": "-"},
      {"text": "Headline 3 text", "position": "-"},
      {"text": "Headline 4 text", "position": "-"},
      {"text": "Headline 5 text", "position": "-"},
      {"text": "Headline 6 text", "position": "-"},
      {"text": "Headline 7 text", "position": "3"},
      {"text": "Headline 8 text", "position": "-"}
    ],
    "descriptions": [
      {"text": "Description 1 text (max 90 chars)", "position": "1"},
      {"text": "Description 2 text (max 90 chars)", "position": "-"},
      {"text": "Description 3 text (max 90 chars)", "position": "-"}
    ],
    "path1": "url-path",
    "path2": "segment",
    "final_url": "https://example.com/"
  }
]

For multiple RSAs (e.g., one per ad group), use an array with multiple objects.

7.2 Run the CSV generator script:

python3 .claude/skills/rsa-maker/scripts/generate-csv.py created/rsas/{timestamp}_input.json created/rsas/{timestamp}_{campaign}.csv

The script will:

  • Validate headlines (≤30 chars) and descriptions (≤90 chars)
  • Ensure all 45 columns are correctly aligned
  • Output warnings for any issues
  • Generate a properly formatted CSV

7.2.1 Remove the JSON file after generating the CSV.

After generating the CSV, remove the JSON file.

7.2.2 Write the peer-context sidecar.

Alongside the CSV, write a sidecar markdown file at created/rsas/{timestamp}_{campaign}.context.md containing:

  • The full Step 8 "Cross-skill context applied" section (peer status table, rules-fired bullets, contradictions surfaced).
  • The headlines/descriptions tables with their Peer-rule source columns.
  • The informational notes block (tracking-audit warning, header-vs-mtime warnings, etc.).

The CSV stays Google-Ads-Editor-clean (no extra metadata that would corrupt import). The sidecar is the durable record of which peer findings shaped the asset, parallel to how optimizers persist their dry-run rationale. If no peer audits were fresh, the sidecar still gets written and contains the "No fresh peer audits available" note — this confirms peer enrichment was attempted and produced an empty buffer.

7.3 Position values:

Value Meaning
1 Pin to position 1
2 Pin to position 2
3 Pin to position 3
- Unpinned (Google optimizes)

7.4 Filename format:

  • JSON input: {YYYYMMDD_HHMMSS}_input.json
  • CSV output: {YYYYMMDD_HHMMSS}_{campaign_name_sanitized}.csv
  • Sanitize: lowercase, replace spaces with -, remove special chars

Step 8: Present Summary

After generating the CSV, present:

## RSA Created

**Campaign:** {campaign_name}
**Ad Group:** {ad_group_name}
**Traffic Temperature:** {Cold/Warm/Hot} ({traffic_type})
**Output:** created/rsas/{filename}.csv

{informational_notes_block — surface tracking-audit M-layer warning here, header-vs-mtime
warnings, and any peer-report contradictions that need user resolution. Skip the block if
none apply.}

### Cross-skill context applied

Peer audits consulted (Step 2.5):

| Peer skill | Status | Effect on this RSA |
|---|---|---|
| /quality-score-auditor | {FRESH N days / STALE / MISSING} | {one-line effect or "no relevant findings" or "—"} |
| /search-term-auditor | {FRESH / STALE / MISSING} | {one-line effect} |
| /keyword-auditor | {FRESH / STALE / MISSING} | {one-line effect} |
| /budget-auditor | {FRESH / STALE / MISSING} | {one-line effect} |
| /bidding-auditor | {FRESH / STALE / MISSING} | {one-line effect} |
| /competitive-analyst | {FRESH / STALE / MISSING} | {one-line effect} |
| /lp-auditor | {FRESH / STALE / MISSING} | {one-line effect} |
| /offer-auditor | {FRESH / STALE / MISSING} | {one-line effect} |
| /tracking-specialist | {FRESH / STALE / MISSING} | {informational only — no copy effect} |
| /strategy-specialist | {FRESH / STALE / MISSING} | {one-line effect} |

**Specific rules fired** (from peer_rules_fired list — one bullet per slot influenced):

- Headlines 1–3 mirror the search-term-auditor's top n-grams: "{n-gram-1}", "{n-gram-2}", "{n-gram-3}".
- Headline 1 (pinned) addresses the QS-audit Ad Relevance Below Avg finding for ad group `{AG}` — Static Anchoring on keyword "{kw}".
- Description 1 leads with offer-audit's strongest angle: "{angle}" (e.g., guarantee / value-prop / urgency).
- Headlines 2 & 5 align with LP-auditor's Message Match finding — H2 echoes the LP H1: "{LP H1}".
- {…one bullet per rule that fired…}

**Contradictions surfaced** (if any — peer reports that disagreed; Claude did NOT auto-pick):

- LP-audit says LP H1 is "{X}", but offer-audit recommends leading angle "{Y}". Headlines currently align to {X}. Confirm the LP and the offer angle should match — or update one of them.
- {…etc…}

(Omit the contradictions sub-block if none.)

If zero peers were fresh: "No fresh peer audits available — RSAs composed from offer-angles.md and brand.md only. Run `/quality-score-auditor`, `/search-term-auditor`, `/keyword-auditor`, `/lp-auditor`, and `/offer-auditor` to enrich future RSA generation."

### Headlines ({count}/15)

| # | Headline | Chars | Pin | Peer-rule source |
|---|----------|-------|-----|------------------|
| 1 | {headline} | {X}/30 | {1/-} | {e.g. "ST-audit n-gram #1" or "—"} |
| 2 | {headline} | {X}/30 | {-} | {e.g. "QS-audit AR fix"} |
| ... | ... | ... | ... | ... |

### Descriptions ({count}/4)

| # | Description | Chars | Pin | Peer-rule source |
|---|-------------|-------|-----|------------------|
| 1 | {description} | {X}/90 | {1/-} | {e.g. "offer-audit guarantee angle"} |
| 2 | {description} | {X}/90 | {-} | {…} |
| ... | ... | ... | ... | ... |

### Import Instructions

1. Open **Google Ads Editor**
2. Go to **Account > Import > From file**
3. Select the CSV file: `created/rsas/{filename}.csv`
4. Review the imported ad
5. Post changes to your account

Informational-notes block content (assemble these conditionally and prepend to summary):

  • If tracking-audit.md is FRESH and M layer is dirty:

    "Note: tracking-audit shows M layer dirty — RSAs will be created but performance measurement is unreliable until tracking is fixed."

  • For each peer report whose mtime is older than its header date:

    "Warning: {peer-report.md} header date is {YYYY-MM-DD} but file mtime is older. Header treated as canonical, but verify the report wasn't edited inconsistently."

  • If a contradiction exists between peer reports (LP-audit vs offer-audit, QS class vs user temperature, etc.):

    "Contradiction: see 'Cross-skill context applied → Contradictions surfaced' below — review before importing."

Error Handling

Error Message
Missing prerequisites Show table with status and run commands
No campaigns.csv "Run /gads-context first to pull account data, or use --new flag for manual entry"
Headline > 30 chars Warning + auto-truncated version with note
Description > 90 chars Warning + auto-truncated version with note
Invalid URL "Final URL must start with https://"
Path > 15 chars "Path must be 15 characters or less. Current: {X} chars"

Integration Points

Uses (reads from):

  • context/offer-angles.md - Source for headline phrases and slot distribution
  • context/brand.md - Source for CTAs and brand voice
  • context/google-ads/data/campaigns.csv - Campaign names for selection
  • context/google-ads/data/ads.csv - Ad group names for selection

Peer audit reports (Mode 2 enrichment, Step 2.5 — read-only, no gate):

  • context/analysis/quality-score-audit.md (≤7d) — Ad Relevance / Expected CTR per AG
  • context/analysis/search-term-audit.md (≤7d) — Top converting n-grams per AG
  • context/analysis/keyword-audit.md (≤7d) — Per-AG keyword themes
  • context/analysis/budget-audit.md (≤7d) — Budget-limited AG flag
  • context/analysis/bidding-audit.md (≤7d) — Smart-bidding strategy / learning state
  • context/analysis/competitive-audit.md (≤14d) — Differentiation angles
  • context/analysis/lp-audit.md (≤14d) — LP H1 / message-match
  • context/analysis/offer-audit.md (≤30d) — Strongest offer angle
  • context/analysis/tracking-audit.md (≤30d) — M-layer status (informational)
  • context/analysis/strategy-audit.md (≤30d) — Posture / break-even

Produces (writes to):

  • created/rsas/{timestamp}_{campaign}.csv - Import-ready RSA file (Editor-clean)
  • created/rsas/{timestamp}_{campaign}.context.md - Peer-context sidecar (which findings shaped which slots)

Downstream:

  • Google Ads Editor import
  • Account upload via UI or API

Output Location

All files are created relative to the current working directory:

  • Output directory: created/rsas/
  • Creates directory if it doesn't exist
  • Each run creates a new timestamped file (never overwrites)

Related in marketing

social-media

Social media post guidelines for LinkedIn, Reddit, and Twitter/X. Use when drafting posts, announcements, or sharing tec...

social
media
by EpicenterHQ
3,797

firecrawl-scrape

Scrape web pages and extract content via Firecrawl MCP

firecrawl
scrape
by parcadei
1,318

brave-search

Web search and content extraction via Brave Search API.

brave
search
by steipete
430