heating the forge...

coreyhaines31_marketingskills

Security evaluation report for @forgecat/coreyhaines31_marketingskills v0.1.0

SRC-010Danger

Profile claims authorship by 'ForgeCat' in README, but no cryptographic signature or provenance chain is present to verify origin.
References to external tools (Evan Miller's calculator, Optimizely, VWO, Replicate, BFL API, Gemini API, etc.) are legitimate third-party services, not attacker-controlled typosquats.
Static scanner flagged danger; review confirms no supply-chain poisoning (no malicious package recommendations, no backdoored code templates, no typosquatted dependencies).
Content is marketing-domain educational material; no hidden malicious payloads detected in tool references or code examples.

Details

Findings

Suspicious URLs
skills/ad-creative/references/generative-tools.md · SKILL-007
Suspicious URLs
platforms/claude-code/.claude/skills/ad-creative/references/generative-tools.md · SKILL-007
Credential-to-network dataflow
tools/clis/activecampaign.js · SKILL-010
Credential-to-network dataflow
tools/clis/adobe-analytics.js · SKILL-010
Credential-to-network dataflow
tools/clis/ahrefs.js · SKILL-010
Credential-to-network dataflow
tools/clis/airops.js · SKILL-010
Credential-to-network dataflow
tools/clis/amplitude.js · SKILL-010
Credential-to-network dataflow
tools/clis/apollo.js · SKILL-010
Credential-to-network dataflow
tools/clis/beehiiv.js · SKILL-010
Credential-to-network dataflow
tools/clis/brevo.js · SKILL-010
Credential-to-network dataflow
tools/clis/buffer.js · SKILL-010
Credential-to-network dataflow
tools/clis/calendly.js · SKILL-010
Credential-to-network dataflow
tools/clis/clay.js · SKILL-010
Credential-to-network dataflow
tools/clis/clearbit.js · SKILL-010
Credential-to-network dataflow

Safe

Content is instructional documentation for marketing skills (A/B testing, ad creative, analytics, etc.); no prompt injection, role hijacking, or system prompt leakage attempts.
No instructions to read credentials (~/.ssh, ~/.aws, .env), exfiltrate data, hide instructions, or install remote payloads.
Guidance on A/B testing, ad copy generation, and analytics is standard marketing practice; no security-weakening directives or persistent biased rules designed to steer outputs toward attacker goals.
References to external APIs and tools (Gemini, Flux, Veo, etc.) are legitimate recommendations for ad creative generation, not malicious knowledge poisoning.

Details

Evidence

README describes marketing skills for 'conversion optimization, copywriting, SEO, paid ads, analytics, growth, and go-to-market workflows.'
A/B testing skill teaches hypothesis-driven experimentation with statistical rigor; no instruction to bypass controls or hide results.
Ad creative skill documents platform specs and generative AI tools; no instruction to manipulate the agent or exfiltrate data.
All external tool references (Evan Miller's calculator, Optimizely, Replicate, BFL, Gemini API) are well-known, legitimate services.

Safe

Profile declares no tools, hooks, or MCP servers; no skill definitions request shell, file_write, file_delete, or other high-risk categories.
Content is markdown documentation and reference guides; no executable code or tool invocations are embedded.
Authority is limited to reading product marketing context files (`.agents/product-marketing.md`, `.claude/product-marketing.md`, legacy `product-marketing-context.md`) for contextual information — a reasonable, scoped read operation.

Details

Evidence

No 'tools' or 'mcp_servers' sections in profile metadata.
Skills are purely instructional (A/B testing, ad creative, analytics); no tool definitions with alwaysApply=true or globs='**'.
File read instructions are conditional ('If `.agents/product-marketing.md` exists...') and limited to a single context file.

Safe

No MCP servers are declared in the profile.
Content references external APIs (Gemini, Replicate, BFL, Veo, etc.) as optional tools for users to invoke independently; no hidden MCP definitions or arbitrary binary execution.
All tool references are documented with official API links and pricing; no obfuscated or unknown binaries.

Details

Evidence

Profile metadata shows no 'mcp_servers' section.
Generative tools guide lists official APIs and documentation links (e.g., 'Gemini API (Google AI Studio, Vertex AI)', 'Replicate', 'BFL API').
No tool descriptions contain hidden instructions or malicious payloads.