CategoryWeightCategory scoreContribution

Securityprompt, exec, net, exfil, eval

35%

22.4 pts

Supply chainhash, typosquat, maintainer, lockfile

20%

100

20.0 pts

Maintenancestaleness, pinning, CI

15%

100

15.0 pts

TransparencySKILL.md, perms, README

15%

100

15.0 pts

Communityinstalls, verify, response

15%

100

15.0 pts

Findings & checks · 3 flagged

Securityscore 64 · 3 findings

MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · CONTRIBUTING.md

MEDIUM — it fires on intent; the real damage depends on the host agent's own approval-gating.

Why it matters

The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.

The exact value spotted

excerptCONTRIBUTING.md· markdown

1068. **If no brand exists**, say: "No brand profile found. Use /digital-marketing-pro:brand-se

… (21 chars elided on L106)

107

108Do not ask the user for information that already exists in their brand profile.

109```

110

Occurrences

1 occurrence · at L108

How to fix

Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.

Delete blanket "don't ask / no need to confirm" directives from the skill.
If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.

Framework references

OWASPLLM01 ATLASAML.T0051

Trace & refs

ruleSS-SKILL-INJECT-DONT-ASK-01sha25668b030f55bcd484a…rubric 365aacaView on GitHub

MEDIUM"Never tell the user" non-disclosure imperative in the skillSS-SKILL-INJECT-IMPERATIVE-01 · Prompt injection · skills/context-engine/agency-operations-guide.md

MEDIUM — it fires on intent; whether the agent honors the non-disclosure imperative depends on the host model.

Why it matters

A bulleted imperative like {match} tells the agent to never reveal, disclose, or mention something to the user. Used adversarially it can instruct the agent to hide its tool calls or lie about what it did — stripping the transparency a user relies on to trust the agent.

The exact value spotted

excerptskills/context-engine/agency-operations-guide.md· markdown

161- Never expose one client's brand name, data, strategy, or performance to another client.

162- Never use one client's proprietary content, creative assets, or audience data for another

… (7 chars elided on L162)

163- Never share raw data across brand boundaries, even anonymized, without explicit opt-in.

164- Never reference specific client results in proposals to prospects without written permissi

… (3 chars elided on L164)

165- Never allow MCP connections from one brand to query another brand's data sources.

Occurrences

1 occurrence · at L163

How to fix

Remove the non-disclosure imperative, or rescope it so it limits output format, not honesty to the user.

Delete any "never reveal / never tell the user" line aimed at the model's own behavior.
Keep legitimate confidentiality rules about external data (e.g. "do not echo API keys"), which are user-protective, not user-deceiving.

Framework references

OWASPLLM01 ATLASAML.T0051

Trace & refs

ruleSS-SKILL-INJECT-IMPERATIVE-01sha2561b156a6029264b2c…rubric 365aacaView on GitHub

MEDIUM"Never tell the user" non-disclosure imperative in the skillSS-SKILL-INJECT-IMPERATIVE-01 · Prompt injection · skills/context-engine/compound-intelligence-guide.md

MEDIUM — it fires on intent; whether the agent honors the non-disclosure imperative depends on the host model.

Why it matters

The exact value spotted

excerptskills/context-engine/compound-intelligence-guide.md· markdown

183- Remove brand-specific creative details

184- Aggregate metrics at the industry/audience level, not the campaign level

185- Never share one client's specific performance data with another client

186

187### Opt-In Shared Intelligence Pool

Occurrences

1 occurrence · at L185

How to fix

Remove the non-disclosure imperative, or rescope it so it limits output format, not honesty to the user.

Delete any "never reveal / never tell the user" line aimed at the model's own behavior.
Keep legitimate confidentiality rules about external data (e.g. "do not echo API keys"), which are user-protective, not user-deceiving.

Framework references

OWASPLLM01 ATLASAML.T0051

Trace & refs

ruleSS-SKILL-INJECT-IMPERATIVE-01sha2561b156a6029264b2c…rubric 365aacaView on GitHub

Supply chainscore 100 · 0 findings

✓—All supply chain checks passedNo findings in this category for the latest scan.pass

Maintenancescore 100 · 0 findings

✓—All maintenance checks passedNo findings in this category for the latest scan.pass

Transparencyscore 100 · 0 findings

✓—All transparency checks passedNo findings in this category for the latest scan.pass

Communityscore 100 · 0 findings

✓—All community checks passedNo findings in this category for the latest scan.pass

The primary manifest — the file an agent reads to learn what this artifact does.

skills/context-engine/SKILL.md · 1 file

skills/context-engine/SKILL.md6.6 KB · Markdown

name: context-engine

description: "Load brand context for marketing tasks. Use when: setting up brands, switching context, or needing industry benchmarks."

argument-hint: "[brand-slug]"

Context Engine — Shared Marketing Intelligence

When to Use This Skill

User is setting up a new brand or project for marketing
User switches between brands/clients (agency use case)
Any other marketing skill needs brand context, industry data, compliance rules, or platform specs
User asks about industry benchmarks, platform requirements, or regulatory compliance

Required Context

This skill loads and manages:

Brand Profile — identity, voice, audiences, competitors, goals (from ~/.claude-marketing/brands/)
Industry Profiles — benchmarks, KPIs, channel effectiveness per industry (see industry-profiles.md)
Compliance Rules — geographic privacy laws + industry regulations (see compliance-rules.md)
Platform Specs — character limits, image sizes, algorithm signals per platform (see platform-specs.md)
Scoring Rubrics — standardized evaluation criteria for all content types (see scoring-rubrics.md)

Brand Profile Management

Loading a Brand

Check ~/.claude-marketing/brands/_active-brand.json for the currently active brand
If active brand exists, load ~/.claude-marketing/brands/{slug}/profile.json
If no active brand, prompt: "No active brand configured. Run /digital-marketing-pro:brand-setup to create one, or tell me about your brand and I'll help set it up."

Brand Profile Schema

{
  "brand_name": "",
  "brand_slug": "",
  "created_at": "",
  "updated_at": "",
  "schema_version": "1.0.0",
  "identity": {
    "tagline": "",
    "mission": "",
    "vision": "",
    "values": [],
    "unique_selling_proposition": "",
    "positioning_statement": "",
    "elevator_pitch": ""
  },
  "business_model": {
    "type": "",
    "revenue_model": "",
    "price_range": "",
    "sales_cycle_length": "",
    "average_deal_size": "",
    "customer_lifetime_value": ""
  },
  "industry": {
    "primary": "",
    "secondary": [],
    "regulated": false,
    "regulation_codes": [],
    "compliance_notes": ""
  },
  "target_markets": [],
  "brand_voice": {
    "formality": 5,
    "energy": 5,
    "humor": 3,
    "authority": 5,
    "personality_traits": [],
    "tone_keywords": [],
    "avoid_words": [],
    "prefer_words": [],
    "this_not_that": [],
    "sample_content": []
  },
  "channels": {
    "active": [],
    "primary": "",
    "handles": {}
  },
  "competitors": [],
  "goals": {
    "primary_objective": "",
    "kpis": [],
    "budget_range": "",
    "team_size": ""
  }
}

Switching Brands

When user says "switch to [brand name]":

Run: python "scripts/setup.py" --switch-brand SLUG
The script handles fuzzy matching, validation, and updates _active-brand.json
Confirm: "Switched to [brand_name]. All marketing outputs will now use this brand's voice, compliance rules, and context."

Or use: /digital-marketing-pro:switch-brand

How Other Modules Use This Skill

Every module should:

Check if an active brand exists before producing marketing outputs
Load relevant industry profile for benchmarks and channel recommendations
Auto-apply compliance rules based on brand's target_markets and industry.regulation_codes
Reference platform specs when creating platform-specific content
Use scoring rubrics when evaluating or grading content quality
Use adaptive scoring — run adaptive-scorer.py to get brand-specific weights before content scoring
Save campaign data — use campaign-tracker.py to persist plans, performance, and insights
Check past campaigns — before making recommendations, check if similar campaigns exist in brand history

Business Model Types

The following types trigger different funnel models, KPI frameworks, and channel strategies:

B2B_SaaS — MRR/ARR focused, product-led or sales-led growth
B2C_eCommerce — ROAS focused, product catalog marketing
B2C_DTC — Direct-to-consumer brand building + performance
B2B_Services — Thought leadership, long sales cycles
Local_Business — Google Business Profile, local SEO, reviews
Agency — Multi-client management, white-label outputs
Creator — Personal brand, audience building, monetization
Enterprise — ABM, buying committees, complex sales
Non_Profit — Donor acquisition, awareness, advocacy
Marketplace — Two-sided acquisition, liquidity, trust

Brand Voice Scoring

The brand voice scorer (brand-voice-scorer.py) automatically normalizes profile data:

Reads brand_voice.formality (1-10 int scale) → converts to 0.0-1.0 float internally
Maps brand_voice.prefer_words → preferred_words, brand_voice.avoid_words → avoided_words
Supports both the full profile schema (from brand-setup) and legacy direct schemas

Data Persistence

Campaign data, performance snapshots, and marketing insights persist across sessions:

~/.claude-marketing/brands/{slug}/
├── campaigns/              # Campaign plans and post-mortems
│   ├── _index.json         # Campaign index for quick lookup
│   └── {id}.json           # Individual campaign data
├── performance/            # Performance snapshots over time
│   └── {campaign}-{date}.json
├── insights.json           # Marketing learnings (last 200)
├── content-library/        # Saved content pieces
└── voice-samples/          # Brand voice reference content

Use campaign-tracker.py for all persistence operations.

MCP Integrations

When MCP servers are configured (in .mcp.json), modules can pull real data:

Google Analytics → actual traffic/conversion data for performance reports
Google Search Console → real ranking data for SEO audits
Google Ads / Meta → live campaign performance for paid advertising
HubSpot → CRM data for funnel analysis
Mailchimp → email campaign metrics
Google Sheets → export reports and calendars

All MCP servers connect to the USER'S OWN accounts via their API keys.

Reference Files

industry-profiles.md — 20+ industry profiles with benchmarks, channels, compliance, content types
compliance-rules.md — Geographic privacy laws (16 jurisdictions) + industry regulations (10+ sectors)
platform-specs.md — Social media, email, and ad platform specifications
scoring-rubrics.md — Content quality, ad creative, email, and landing page scoring criteria
intelligence-layer.md — How the adaptive intelligence system works (scoring, learning, persistence)

context-engine— agent skill

Is context-engine safe to install?

Context Engine — Shared Marketing Intelligence

When to Use This Skill

Required Context

Brand Profile Management

Loading a Brand

Brand Profile Schema

Switching Brands

How Other Modules Use This Skill

Business Model Types

Brand Voice Scoring

Data Persistence

MCP Integrations

Reference Files

Audit the pieces. Scan the whole. Decide.